Today, I’m talking with Benn Jordan, musician, YouTuber, researcher, and investigative journalist. You might know Benn from those first two titles — he’s got a prolific discography of experimental music dating back decades.

But in the last couple of years, he’s become a high-profile face in the creator and media spaces for his work investigating Flock Safety, the maker of automatic license plate readers and the software that powers them.

Benn recently testified in front of Congress, at a hearing the Senate Judiciary Committee’s Subcommittee on Crime and Counterterrorism held, dedicated to Flock and other ALPR systems. This hearing felt like the culmination of months and months of growing criticism and bipartisan backlash against Flock — all of which is taking a shape similar to the pushback against AI data centers.

Verge subscribers, don’t forget you get exclusive access to ad-free Decoder wherever you get your podcasts. Head here. Not a subscriber? You can sign up here.

This has led to major fallout for Flock. Hundreds of cities, counties, and other local governments have begun canceling contracts, and Flock is now under investigation by Republican Sen. Josh Hawley. Last week, Reuters reported that Flock plans to cut about 18 percent of its workforce.

So of course I wanted to dig into all of that with Benn, not just because he played a role in that, but because he’s also become in some ways the face of a new front in the broader debate about surveillance and privacy.

A point you’ll hear Benn make is that he doesn’t actually hate Flock as a company — he thinks its posture toward critics has been a colossal PR failure that made it low-hanging fruit, while competitors like Axon and Motorola quietly build the same kind of networks. Instead, Benn now talks to Flock directly, including to its CEO Garrett Langley, and reports security vulnerabilities to the company when he finds them. (As a side note: we have also invited Flock’s CEO on the show; Garrett if you’re listening, we would love to have you.)

Benn says his outreach to the company is an olive branch he’s extended to Flock as part of his goal of working toward real, tangible solutions to surveillance overreach, rather than part of a crusade to tear one single company down.

That’s a real theme throughout our conversation: that Flock and other high profile names in surveillance are just part of the problem — our real focus should be on the interlocking relationship between private industry, government, and law enforcement that prioritizes secrecy over transparency.

We never voted for an insecure network of cameras on every street corner, and we’re certainly not being asked if we’re comfortable with access to this network being given to police officers with a penchant for abusing it. So Benn sees his work on YouTube and with other researchers, civil rights groups, and journalists as an accountability function that forces this conversation to the forefront. And so far, it’s working.

Okay: Benn Jordan on Flock, privacy, and the new surveillance backlash. Here we go.

This interview has been lightly edited for length and clarity.

Benn Jordan, you’re a musician, YouTuber, and researcher. Lately you’ve been an investigative journalist. You are a witness in Senate hearings. Welcome to Decoder.

Thanks for having me.

I’m a huge fan of yours. It’s just a delight to have you here. A lot of people in the Verge audience know you as a musician, but lately you have been relentlessly focused on Flock cameras, which are automatic license-plate readers (ALPRs). You recently testified in front of a Senate hearing on those cameras and the surveillance they enable.

Give us a sense of your background. For people who don’t know you or haven’t been following you for years, where did Benn Jordan come from and how did you end up being the person who makes videos about Flock?

I’ve been a musician my entire life, pretty much. I’ve also always been kind of a hacker, or somebody who is that type of person who always takes stuff apart. If I get a product that wants me to pay for a service that’s already in the product, then that’s a puzzle I want to solve and I get excited about it. But a lot of times that blends into investigative journalism as well.

With the Flock stuff, it’s just that I live in Atlanta, which has more Flock cameras than anywhere else per square mile. About two years ago I started noticing them. I thought they were traffic cameras or something like that. Then I started looking into what they were actually doing and then started looking into things like security vulnerabilities. More than anything, everything that’s happening right now in the public, in the news, has to do with people just figuring out that these have existed the entire time.

A lot of my focus now is on the lack of transparency from the government, because the companies are always going to make money and try to make as much money as they can, especially if it’s from taxpayer money. But in this case, I feel like our government missed it. And that’s where we’re at.

There’s a question that’s been rattling around my head the entire time we’ve been having what I would call the modern privacy and surveillance debate.

When I say “modern,” what I mean is it feels like we woke up sometime in the last year and everything is recording us all the time and everything is watching us and the glasses have cameras in them and Mark Zuckerberg is saying phones don’t have a light, and that is meant to end a conversation.

That’s it. That’s the argument. Phones don’t have a light, no further questions. The Apple Watch is going to start doing automatic transcriptions, and everyone said, “Wait, hold on. No thank you. We actually don’t want this to happen.” But my personal experience of this is that my political coming of age was the Iraq War and the Patriot Act.

This is the stuff that people were talking about twenty-five years ago — don’t do warrantless wiretaps, don’t make the databases all connect to each other so you can take the metadata and de-anonymize people at will.

What is it about this time that has made everybody so angry? Because we could have been angry twenty-five years ago.

I’m always trying to get people more informed on the information that’s being collected about them every time they go to a website, every time they sign up for something, every time they check into a restaurant online, every time they use their phone. In this case, one of the reasons that people are really upset about is that you can’t opt out. That’s the first thing.

I can’t say, “I don’t want these Flock cameras or Axon cameras or whatever they are tracking my every move and giving it to law enforcement.” I have no option to do that. I was never given an option on whether to vote for them or vote for somebody who was in favor of them. It just kind of happened and you know people weren’t told about them. That would be the first thing — that you can still install a separate OS on an Android phone or you could turn off Wi-Fi and Bluetooth or just leave your phone at home. Whereas now you can’t actually go to the grocery store without somebody tracking your vehicle.

The second thing is that — and this is kind of related to opting in and opting out — I can pay for a new AI smartwatch or I can pay for an Amazon Echo device or something like that, or I cannot. Whereas this way I’m paying for this already. This is just my property taxes going up and I’m paying for surveillance.

In some ways it is the same as the NSA, but I feel like it’s much more local now.

You don’t get to go to the town board meeting and yell at the NSA. Although maybe you should. Honestly, that might be good for everyone. There is that part of it, and you’ve certainly been at a bunch of local town board meetings. I’ve watched a lot of videos of you talking to them and helping other people articulate their concerns.

That, to me, is one of the more fascinating aspects of all this. Flock went from a company that was selling cameras to everyone, and the idea that adding Flock cameras would increase the safety of your town went from all but assumed to, “Hold on, we all have a lot of questions,” and now this company is struggling a little bit. They’re losing contracts, they are retrenching, their communication strategy has changed.

What do you make of that? Is that just the power of local democracy in action, or is there something else?

It definitely is democracy in action. That’s the good news. The best news out of anything that’s happened is that with both Flock and data centers, it’s two things that have made people actually look up who their mayor is and who their city council members are and show up to meetings and then get involved in those local politics.

Because those local politics affect your life way more than national politics. They just don’t make the headlines. We’re not as involved. Your day-to-day following what Donald Trump is doing is likely not going to make that big of a difference other than what you argue about over Thanksgiving. Whereas local politics, that’s straight up, in the next year, going to have a big effect on your life.

People are getting involved in that, and then in cases like Asheville, North Carolina, where it’s actually changing the political landscape. People are so dissatisfied with the surveillance policies of their local politicians that it looks like that they may have a new mayor as a result of it, who also comes with a bunch of other interesting and progressive ideas that would probably help the city a lot more than the current mayor has.

That’s a big deal right there. The other part that discourages me a little bit, or the part that I find a little disappointing, is that Flock has become this buzzword, but a lot of people don’t look beyond that. I just heard in the race for governor in Georgia that the Democratic candidate yesterday talked about getting rid of Flock cameras and being anti-Flock, when this is somebody who as mayor of Atlanta welcomed Flock. You see a lot of that happening now, too, where they just say “Flock bad” and then they go sign an Axon contract or they put Verkada or Motorola cameras everywhere.

In my research, Flock has been the low-hanging fruit, just because they were the loudest and most defensive, whereas every other company was just taking a step back and cashing the checks.

One of the pieces of the puzzle here that is just fascinating is that Flock is a startup, essentially. They’re a new company, they have a new go-to-market strategy. They talk a big game about what AI can and can’t do and with what they as a platform should and should not enable. Your Axons and your Motorolas are government contractors. They are not companies that are meant to have a personality.

There’s something about Flock where Garrett Langley, the CEO, had a personality. He would go on TV and deride any critics as left-wing terrorists and that did not work. There’s something about surveillance in that case where that strategy didn’t work. There’s something about data centers where that strategy didn’t work.

We just ran a big piece about the people of Utah pushing back on Kevin O’Leary, who called them all Chinese agents, and they were like, “What are you talking about?” That cut across party lines. Again, you’ve been out there. I know you’ve talked to a lot of people pushing back on surveillance.

Is it just that Flock had a personality, and that enabled bipartisan pushback? Or is it surveillance more generally?

In terms of the media getting attention, Flock had a lot of personality and they were very defensive. I’ve met a lot of the people from Flock who are so villainized and they don’t seem at all the way that they’ve seemed in the clips that I’ve covered in my content. It’s very different. You would assume that there’s a publicist behind a lot of this.

Especially in something that serves law enforcement, where you would maybe have a publicist that’s used to dealing with Republicans and right-wing things, they just tell you to keep repeating the same thing over and over again and take a firm stance and dig your heels in. Most people who aren’t Donald Trump, when they do that, get wrecked. Flock got wrecked. They kept saying things and they kept denying security vulnerabilities and they used clever ways of reframing or redefining words and people picked up on it and now they are the villain.

Whereas companies like Axon are not as sexy for a journalist. They just aren’t, and that’s probably intentional on their part. Any company that’s taking your tax dollars or trying to court your government to give your tax dollars to them probably wants to be out of the spotlight.

We’ve covered Axon before. They do not want to be a company with characters. They are as anonymous as they can be.

One of the things that I want to see with you on the show, most importantly of all, is to describe what a Flock camera is and what the technology behind the hardware does that makes it new.

Because from watching all the videos that you make, I think, “It’s just an Android phone. They’re just putting a bunch of Android phones around town,” but it’s actually the database and the AI that makes it more dangerous than usual. Describe what the actual Flock hardware is and then how that connects to the system that makes it as controversial as it is.

So ALPRs have existed, I believe, since the 1970s. They started getting used in the United States in the late 80s and 90s. One of the classic ways that they would be used in the 2000s is they would be mounted on a police car. If a police car wanted to run someone’s plate, that way they didn’t have to type it in while they’re driving, and they could run a plate over to dispatch. I personally don’t have any problem with that. I think that makes sense. That’s a decent use. However, you would have issues where you would get plate misreads and then people would get taken out of their car at gunpoint as a result.

Flock is a lot different. Flock basically started a private nationwide network where any police department or city that’s using Flock’s network can opt into it and share their data with it. Police departments across the country can then run a license plate with 30 days of retention on everywhere someone’s been, anywhere in the country, as long as they’re sharing the data. So you would have someone in Texas being able to search somebody’s movement through New York or Atlanta or something like that.

They were doing this — they are doing this — without a warrant. Whereas any type of ALPR data sharing previous to Flock had to go through the National Crime Information Center, which does require a magistrate judge to look at the overall context of a situation and then issue a warrant to allow tracking.

That’s the innovation here. They built the private network, the private database, and convinced a bunch of law enforcement agencies to buy their system and they could opt in. There’s some kind of legal innovation here where by making it a private database, they could just excuse themselves from the warrant process.

But the actual hardware of the cameras is not all that impressive. There’s something about that that I really just want to hone in on. The idea of the Flock camera being the problem is secondary to the Flock network being the problem.

One way that I think about it, in the defense of Flock cameras and even outside of Flock itself, is that it’s all happening in public. This is all in public view. It’s generally the same as arguing that you’re allowed to take a picture anywhere in public. So why can’t they do this?

My argument against that would be if you have a police officer follow someone or go through public traffic cams, which are all over the country, and go through every single public traffic cam and wait for your car to come and then write down when that was there and do that for 30 days, that’s generally a serial killer situation.

The amount of resources that you have to dump into that would have to justify that expense with a very serious crime. But doing that for somebody who stole steaks from a Kroger is very different. What worries me about it is that you might end up with police officers being able to get a lead on a case much easier that doesn’t necessarily need police for society to become safer in a better place, like a nonviolent crime or something that’s not a huge deal, rather than investigating more serious crimes.

That’s the argument that Flock makes, and that proponents of this technology make. We had Jamie Siminoff from Ring on the show, and the argument he would make is if you put enough cameras into certain neighborhoods, you will “zero out crime.” That is a thing that he has said to me in the context of Ring doorbells and floodlight cams.

It’s the same argument that proponents of Flock make. “We put up all these cameras, we can catch criminals, thus we will deter criminals, the crime rate will fall.” Do you buy into this?

No, I don’t. Because first of all, we’ve never tested this. We’ve never had a city in a vacuum where we put in a bunch of cameras. One of the metaphors I like to use is if you have a supermarket and you put one camera in one aisle where you think that things are going to get stolen, that will probably increase theft in other aisles rather than just assuming that every shoplifter is just going to become a plumber instead, like I’ll be a doctor instead of a shoplifter.

That doesn’t actually deal with the overall problem. If you have a Walmart that has cameras all over it, and then you have a neighborhood store that can’t afford all of those fancy cameras, then maybe the shoplifters and people who would hold up a store will go to that place instead and you have the same thing with cities. You have crime relocation happening.

So, yes, of course, if you put a camera above a street, it’s less likely that a car on that street within the view of the camera will get broken into. But does that make it more likely that a car a block over where there is no camera will get broken into? This is something that has been observable by criminologists, like crime relocation is a real thing. A lot of times effective policing just moves criminals elsewhere.

Selfishly, you could say, “In my subdivision, we have less crime, and I don’t care about anything other than that.” But over time, over the next decade, it doesn’t really work out that way. You end up with more crime overall, or you end up with bigger problems in society. So I would argue with that.

Also, we’ve seen that Ring cameras don’t really reduce crime all that much. You end up being able to see somebody stealing a package off your porch or you end up being able to see somebody breaking into your house. But do they actually increase clearance rates? I would argue I’ve not seen any evidence that they do.

You have to connect them to the nationwide database of license plate readers.

[Laughs] Right, yeah.

If you put a camera everywhere, you can’t relocate anywhere. That’s the argument. It’s the end point of the argument that they are making: “If putting up some cameras would move the crime, putting cameras everywhere would make that impossible and there will be no crime.”

With a lot of that stuff, it’s just like, “If you have a problem, if you’re worried that you’re going to have a porch pirate, if you’re worried that your house is going to be broken into, I’m not sure why you would jump to put cameras all over your house?”

It’s the same thing with cities, like why wouldn’t you just increase your insurance? You know, increase your insurance payoff or something like that, because that actually will get you new things and you know, the other one will just get you a police report with evidence that the police might not even be able to handle or will not be able to connect to any sort of criminal.

Because even if that criminal gets caught, it’s not like they’re just saving all the money to pay back their victims. It’s figuring stuff like that out. Another thing that I don’t like about surveillance in general, especially things like ALPRs, is that they’re not solving crime with embezzlement. They’re not solving crime with corruption. They’re not solving bigger crimes.

They’re really honing in on things like car theft and things like shoplifting out of a Lowe’s parking lot and then sharing that data with the police and then the police tracking the car two days later and pulling it over. Most of the time those crimes actually have solutions that are in broader public policy.

If somebody’s shoplifting from Lowe’s, I would assume they probably either have a drug problem or they have a problem with poverty, because I don’t want to shoplift from Lowe’s. I don’t have any desire to do that. It’s not a crime of passion. It’s a crime of necessity. I would say that like along the lines, like even if the clearance rates were improving, which I’ve not seen evidence of, what does that mean? Does that just mean that we’re putting more nonviolent offenders in prison or through the justice system? Does that actually make society safer?

We’re getting into the trade-offs of using technology to solve social problems. One of the most important trade-offs that you have brought to light repeatedly is that these devices are not secure.

So even if you solve the problem of “you should get a warrant” — or even if Flock responds to the public pressure and says, “We ought to make people type case numbers into the system before they start searching,” which was their big concession recently — even if you get to, “We’re going to control the system a little bit better,” the devices themselves are insecure. The network is insecure.

That means anybody can do almost anything they want with it. You’re a security researcher. Describe how you began digging into the security problems here and what the response has been.

Starting out, I found out that they didn’t, in early 2025, require multi-factor authentication. I also witnessed a police officer logging into Flock from a Wi-Fi network that wasn’t at a police station. Had you told me those things, I wouldn’t believe them. But witnessing it happening made me say, “This is something that needs more looking into.” I started thinking, “It’s probable at that point, because now you have human error.”

If all of these police departments are sharing data with one another, then it could just be the dispatcher, somebody who’s not even deputized, who has a Flock account at a rural Louisiana police department. That person could sell their account or that person could click the wrong link in an email and whatever. I started looking for Flock credentials on the dark web and I found some and they were coming from a Russian vendor.

I started thinking, “How exactly did that happen?” That’s how I met [cybersecurity researcher] Jon Gaines, and we found that a lot of the time there were hardware vulnerabilities in the camera where you could get them to open up a Wi-Fi access point, and then you could get complete control over the Android device and you could root it and have it do whatever you wanted in a way that would be undetectable because it’s happening outside of the Flock ecosystem on the device.

I had found a bunch of Flock cameras with administrative interfaces that were open on the internet. So you could just type in the IP address and then delete footage, or watch footage from 30 days ago. You could even put files on the device and look at all the logs. In all these situations, we found a lot of plain text credentials, things that shouldn’t have been there.

At first the pushback from the company was that I’m a YouTuber and I was doing magic tricks and I’m not a serious security researcher. Reading their own statement about having never been hacked through a Flock camera and then downloading it on my own machine made that indisputable.

They still disputed it for a while, but that is what really got the media involved and finally paying attention to it. Since then, it’s been smaller things that I’ve found, but more so the bigger deal here is that no matter if you’re using Flock or Axon or whatever it is — because now you have Axon and Motorola also launching their own private networks outside of the National Crime Information Center — the person on this network with the worst security hygiene is going to be as secure as the entire network will be. That’s the big problem.

If I were to have it my way, police officers would not have access to Flock at all. If I were to have it my way, you wouldn’t have this level of surveillance at all. But let’s just say if I were to have it my way on security policy, a police department would have one person who is trained and has a lot of security hygiene and privacy hygiene and is trained to look up Flock searches for police officers. But what we’re seeing is a lot of police chiefs who have Flock accounts. Why would a police chief ever even need to use this service? It’s their job to manage the department.

In the action movie, the cops usually have to bribe the person with access to the surveillance tool with cookies and booze to run a search without a warrant, right?

Yeah, and they usually have bifocals and are really smart.

We’ve been conditioned to think that’s how it should work. At the very least, there should be some light bribery involved, not free access. I am not a security expert by any means. I’ve been covering tech for a long time, and it just strikes me that there are some very basic concepts about security that Flock in particular missed.

The idea that we have open access, there are no access controls, that they were practicing what is called security by obscurity. You could just drive a truck through it and they would deny it. All that seems very basic. Is it just that they’re a startup and they didn’t know any better?

I think they knew better and they didn’t have to do anything about it, because nobody made them. Because somehow across America, police departments and city councils and senators and your federal government allowed these cameras to be put up everywhere without any type of security test or any type of security protocol, any type of triage for what to do if something goes wrong. None of that exists.

Circling back, when I’m actually sitting there in front of the Senate, I’m not going to say, “I can’t believe how bad Flock is.” Flock is a for-profit tech startup. Of course they’re going to do whatever they can to stay competitive. That’s their job that we give them as a society. The job of a legislator is to protect us from them. Because even when we yell at police for the privacy stuff, you wouldn’t consider a police officer or a police department to be Fourth Amendment defenders. They would be on the other side of that argument.

We need somebody on the side that actually defends people’s privacy. Our government has been utterly asleep at the wheel, especially when it comes to security vulnerabilities, but that’s across the board. That’s everywhere. Right now, it’s actually a really terrifying time that we’re in.

You recently testified in front of the Senate Judiciary Subcommittee on Crime and Counterterrorism. That was a hearing about Flock. That subcommittee is chaired by Josh Hawley, who’s a Republican. He opened a probe into Flock.

I’m curious about that experience, preparing for that testimony and what message you wanted to send. But I just want to note that it is remarkable that a Republican is holding hearings into law enforcement technology and whether or not it’s appropriate, and that’s happening across the country.

What do you make of that? That it’s this bipartisan? Is it that everyone’s just woken up to civil liberties? This is a convenient punching bag, and everyone hates tech? How do you see the shape of it?

For any politician at the Senate level, there’s going to be theater. I don’t consider that to be a meeting where there’s going to be a solution at the end.

But what I did do there, in full transparency, is I wanted to testify with the C-SPAN cameras on in front of a Senate committee, explaining that things like FISA 702 or CORCA — these things that a lot of these politicians, even some people on that subcommittee, co-sponsored — are massive risks to privacy and security.

Because you can’t say that they didn’t know when an expert witness who’s a hacker told them in testimony. That’s ultimately why I went to Washington, DC, because I knew that it would be basically just a bunch of people proudly saying that they’re angry at Flock because everybody’s angry at Flock.

A lot of the things in Hawley’s proposals are things that Flock is already kind of doing, in their own defense, which is good news. It’s not to choose one over the other. Both are actually taking positions that are big compromises from where they probably stood a year or two years ago.

It was a wild experience. It was strange. They expected me to just talk about how bad Flock was. I really didn’t really see a purpose to that. If I wanted to scream about how bad Flock is now, I would just contact somebody at Flock and say, “We need to talk about this. I’m really mad about the security vulnerability,” because they do talk to me now. If I find a security vulnerability, I tell them about it and they take it seriously. Of course it took a lot of damage for them to take me seriously.

But if I’m standing in front of the Senate, I’m going to say, “Why don’t you do something? Because this is your job and this is what we voted for you to do.” As far as the bipartisan stuff, I’m really happy about it, honestly. It’s great. It’s awesome to see our country agree on anything as a whole right now. That’s wonderful.

A lot of Republicans remember COVID and they remember mask mandates and stuff and they probably imagine what a Flock camera would do in that situation. It might automatically send them a ticket or an arrest warrant for leaving their house during lockdown. Who knows? But that is all related to privacy risks and authoritarianism and it’s all very real in their memory.

It is interesting to see a lot of politicians take the side of ICE and take the side of the National Guard driving into a city where the police department doesn’t even want them there, and then also be anti-Flock or anti-surveillance. It’s an interesting time.

There’s something there. One of my early career rookie mistakes was that I covered telecom companies and ISPs as a baby blogger at AOL and I got it into my head that I should testify in front of Congress to talk about how the ISPs were charging everybody too much money.

I called the general counsel and the policy office at AOL and I said, “Can you get me in front of Congress?” They said, “About what?” And I said, “The ISPs are bad,” and they went, “You know, we’re an ISP,” and that was the end of that conversation. It didn’t go any farther.

[Laughs] You definitely need a lawyer if you’re going to testify in front of the Senate. There are a lot of legal expenses involved. It’s pretty stressful all in all. I feel like a lot of people think it’s a lucky thing, but  it’s a very stressful two weeks of your life.

You are in the center of that swirl, and it is bipartisan. I would put up surveillance and data centers as the issues that are uniting Americans today. Are you hearing from both sides? Did you sit with Republican senators individually? I’m sure you hear from regular people on both sides of the aisle.

Is it really that we don’t want the government to spy on us, so we’re going to yell at these cameras? Or is it bigger than that?

I think, and what I hope, is that when people discover what’s going on with Flock, and what’s going on with other police surveillance technology, they also don’t immediately go to misinformation, because that’s what a lot of people do. A lot of people will hear about the Flock cameras and then they’ll go, “How is Israel involved in this?” Then they immediately go down that rabbit hole on X or 4chan and then it builds this conspiracy theory that’s completely not in parity with the actual problem that we’re trying to solve.

What always entertains me so much about those types of conspiracies is that reality is just as crazy. It’s just as wild. There’s just as much outrage in reality. Somehow conspiracies are more attractive, even though on many occasions they’re not even as interesting as reality.

LexisNexis, for example, is a data broker that police use, that your car uses, that your devices sell data to constantly. If people knew the average amount of data that this data broker has about them that’s being sold and traded constantly, I feel like their anger about Flock would be completely eclipsed by it.

Can I connect those thoughts? I’ve been thinking about this a lot. Five years ago, we could confidently write a story debunking the notion that Instagram was listening to you. [Instagram CEO] Adam Mosseri, this poor man, has to make a video once a week insisting that he is not listening to you. Part of the reason that five years ago we would be confident in writing that story is the technological cost of listening to every Instagram user and figuring out what they had said to serve you ads was actually very high.

It was much cheaper to say, “You were on your friend’s Wi-Fi network six months ago, they searched for vacations, we’re going to serve you the ads and the coincidence will be astounding and you will believe that we’re listening to you, but it’s actually just a bunch of like data matrix synthesis, right? But that’s going to be very effective.”

We could write that story and say, “Actually you were in the same place at the same time. There’s no way they’re listening to you, it would be too expensive.” But here in 2026, I’m not sure I can make that claim anymore. Everyone wants to listen to you all the time. The amount of databases that we are synthesizing have skyrocketed all the way to “I’ve seen all the license plates on all the highways at all times, and the police chief has that data.”

How do you regain a sense of privacy? There’s a part of me that says so many people believe their phones are listening to them that you can never claw your way back, that you can never end the conspiracy theory.

The “phone listening to you” thing is so interesting to me because it’s one of those perfect examples of conspiracy versus reality, where if somebody is having a discussion about something, like that they want to get a horse. Weird example. But I was thinking about learning how to ride a horse. You have that discussion with your partner and then later on that day, you get a suggestion or an ad for horse-riding lessons on your phone. The conclusion and the conspiracy theory is my phone is listening to my conversations.

That’s pretty reasonable until you find out that it’s very inefficient for them to do so. The crazy part is that you’ve been giving subtle hints to Meta, to Google, and to all these companies and they have an algorithm that can figure out that you’re going to want horse-riding lessons and so when you feel like your phone’s been recording your conversations, it’s almost like this glitch in the matrix where you have a moment to realize how powerful the algorithm actually is at predicting what you’re about to think. That is scary and that’s not a conspiracy. That’s actually how it works.

That always boggles my mind like that. Wow, look at the real thing happening because it’s so much crazier than the conspiracy is. Of course they could listen to you. There are a lot of devices that listen to you that say they’re using your data to self-improve. Tesla listens to you. Like there’s a lot of different services and products.

But the scary part is that listening to your conversation is one of the least efficient and least valuable things it could do because it already has so much more important information. What you’re communicating verbally is usually not nearly as valuable as just your subtle pattern.

Yeah, and your transaction history. I guess what I’m getting at is again, five years ago, I would say some version of that all the time: “They’re not listening to you. The real thing is way scarier and listening to you isn’t valuable.” And it would be costly. You have to store all this audio, you have to process it, you have to figure it out. And now it feels like actually it’s not costly. It is happening all around us all the time.

If I look at Flock, the problems break down into two buckets. One, the very idea of a private database that anyone can search without controls is bad. And we should legislate a bunch of controls around that thing. And then two, letting a startup kind of YOLO that technology into the world without any security best practices, equally bad. These are two different problems that have come together to create a privacy disaster.

But even if you think Axon or Motorola will have better security, you’ve got the first problem, which is you need controls around these private networks. All that in my mind is right next to the conspiracy style thinking you’re describing, where it’s technologically plausible that all the bad things are happening.

I can’t rule them out anymore. My Apple Watch might just make transcripts of all the meetings without anyone’s permission. The loophole you describe in our legal system for years, the idea that you don’t have privacy in public, has been a lodestar. It is just foundational, and now people don’t love that anymore. Because the idea that you don’t have privacy in public enables a bunch of people to do a bunch of things people don’t like.

Maybe we should rethink that as a moral, ethical, and legal matter, but we’re nowhere close to that. I just feel like all that is pouring fuel on the conspiracy theories. You’re in the center of the storm here, and so I’m wondering if you can find a rational set of policy outcomes here that tamps down on those flames.

A little over a year ago, when my first video on Flock came out, if the company’s response and the government’s response to it was, “Audit logs are now going to be open and they’re going to exist on something that uses immutable data storage. It could be the blockchain, could be WORM or something like that. Basically, anybody can look at them, anybody can own them, nobody can change them…”

If they said that, then they immediately went and fixed all the security vulnerabilities and you just had a smarter way of looking at these things… Also one thing that would have been really interesting around them is to be able to go to your DMV and show them your license and opt out.

You could say, “I’ve not been convicted of a felony, I’ve not done anything wrong, I don’t want to be tracked.” And then all of a sudden, when a Flock license plate reader detects your plate, it says “opt out,” unless a cop puts you on a hot list because you’ve murdered someone or something like that. These are all reasonable compromises that could have been made with relatively little expense.

But now a year later, who’s going to trust anybody who makes those compromises? Now there’s no trust anymore because of that. Because Flock dug its heels in so much, and because police departments constantly gaslit everybody over and over again with, “There’s no abuse, there’s no record of abuse. We’re checking it. We’re checking these audit logs constantly. It makes you safer. If you take these cameras down, your kids are going to go missing.” That type of thing.

Now how can you fix this? Now people are just destroying cameras. Now, people are just threatening executives of these companies because that’s what they do when they get taken this far out.

That’s my big problem now. If I have any goal in this or any dog in this fight, I’d like to bring people back towards solutionism, because I don’t want surveillance at all, but we don’t live in Benn Jordistan. We live in a democracy and a lot of people are okay with it to some degree. Most people are not okay with it, from polling I’ve done, if it doesn’t involve a warrant.

Somewhere along there is a solution where nobody’s happy, but everybody’s content. If I could do anything right now, it would be bringing people back towards a path where a solution can be found.

One of the things I’ve been thinking about a lot is the widespread assumption that surveillance does exist. When there’s some high profile crime, the assumption that there will be cameras that will show us something about this crime actually kind of redounds in favor of the surveillance companies. Ring had an ad at the Super Bowl about tracking lost dogs. They got in a lot of trouble over what else we’re going to do with this surveillance network.

I think it was a week later that Savannah Guthrie’s mom was kidnapped and Jamie was on every news network in town saying Ring cameras are going to solve it, whether or not the footage that we saw was even from Ring cameras.

There was this notion that surveillance would help. That is a real pendulum swing back and forth. The cameras are the solution at one time and the abstract concept of the cameras makes people mad because why are you passively surveilling me? You’re talking about solutions. How do you thread that needle?

Most people need to understand, especially police departments, that most of the things that are being promised by people like Garrett Langley, the CEO of Flock, and Jamie Siminoff, the CEO of Ring, are their sales tactics. These CEOs or executives or anybody from any company, they’re not actually going out and doing this research in a way where they’re not cherry-picking data and looking for statistics that support their product.

You could say that’s unethical, you could say that’s wrong, but that’s what you expect from a tech company. Nobody believes that a tech company would all of a sudden become a humanitarian while they’re still representing their company. We could take that and we could isolate that, and then we say, “Research needs to be done,” and research is very, very complicated for this.

In sociology and criminology, we don’t even understand if increasing police reduces crime as a society. We’ve not even proven that. Putting cameras up is a really big ask. We just need to be asking these questions a lot more, and being more careful about putting these cameras up. What will they actually help?

If we could find something that actually helps, we could say, “We have a lot of cars being stolen in this neighborhood. Hey guys, how do you feel about an ALPR that has 24 hours of retention time coming up in this neighborhood and everything that we search on it will be in an immutable audit log that you’ll be able to see every single day? How do you feel about that?” I feel like a lot more people would say, “Yeah, I don’t want my car stolen.” They would maybe err on the side of trying something to reduce a crime problem rather than the police just behind closed doors signing a multi-million dollar contract to put cameras up all over the city without asking anyone.

That’s one of the first solutions. Transparency in general. If I were running Flock in some weird parallel universe, a year ago when Have I Been Flocked? came out — the website that allows you to search your own license plate — I would have immediately partnered with them. I would have said, “Here’s all our data. Thank you for running this. Let’s give you some money so you can keep this server running fast. Because I want everybody to not be mad at us. I want everybody to understand what’s going on.”

A lot of it comes down to quantifying things at a local level to see what people want, and maybe even having referendums in certain places. Referendums are usually really tricky and you have your Brexit situations with referendums, but sometimes they’re also useful.

For a general person watching this right now, the one thing that I strongly urge is to find out who’s voting for this stuff. Find out who actually signed the paperwork to spend the money on it without any form of security vulnerability check, without any sort of special license or any sort of audit data, without going through anything else, because that person is typically a city council member and you can run and be one of those quite easily.

It’s not that difficult. Most city council members that I’ve met are just like me, just like my viewers. They’re not these prestigious politicians. It’s something that most people could do if they have a few friends helping them campaign.

You said immutable record again and it reminded me that earlier you earned the distinction of being the only person on Decoder to ever suggest the blockchain as a solution for something in a way that did not immediately imply that you would make a bunch of money. I want to congratulate you for that.

[Laughs] It’s interesting that all of this ALPR stuff blew up at the same time that crypto and blockchain and NFTs blew up. Yet nobody ever thought, “Hey, we could use a proof-of-stake network, one that’s not actually scorching the Earth to solve Sudoku puzzles, but we can actually use a highly efficient proof-of-stake network to immutably store data for these audit logs.”

But it’s not too late. These things still exist. You don’t have to use blockchain. There’s IPFS, which is a decentralized file system. There’s WORM, which is “write once, read many.” That’s a thing that IBM, I think, invented, but it’s used everywhere. There are a lot of ways to do immutable data, but it being democratized data is what’s the most important part.

[Laughs] We’re not using the GPUs for Sudoku puzzles anymore. We’re running them red hot for other stuff. Don’t worry. We switched those over.

You mentioned working with the companies. You mentioned the idea that Flock is now partnering with some of its critics, which would have helped them early. They have invited you in, right? One of your strongest calls is for external auditors and researchers to evaluate these systems. It’s one of the things you said was needed in your congressional hearing.

They’ve invited you in. You’re now a third-party reviewer with Flock. What is your relationship with Flock like now? And do you think it’s changed how you view and evaluate the company?

There are certainly people at the company who really dislike me. And then there are people at the company… I’ve spoken to the CEO. I’ve spoken to a couple of the executives. There are people at the company who are absolutely willing to have a conversation and figure out what a solution actually looks like. Unfortunately, and I say “unfortunately” because throwing poop is really easy, but cleaning up the room so people can sit in it again without smelling poop is really hard, and most people are only willing to do the throwing poop part.

But one thing that happens in our society all too often is when we have a big controversy, especially on social media, and then there is no remediation at all. There’s no allowance for remediation. If you think about somebody getting canceled or something, one thing I always think is, “What do you want them to do? What should they do now? Do you have any ideas or should they just die?” Because that’s not a solution.

We can’t just snap our fingers. If, in the next month, Flock just said, “We can’t do it anymore. We’re going out of business. We quit. We’re going to make toys or something else, or we’re going to just leave,” Axon, Verkada, Palantir, and all these other companies would just step in and do the exact same thing and they would learn from Flock’s mistakes and do it more efficiently.

We’re in a place right now where, had Flock not dug their heels in and not listened to press strategies that make them seem more confident than they actually are, and actually had listened to people and maybe worked on a solution — it’s going to be rough and I’m going to get a lot of hate for even suggesting that that’s a thing — but if we actually want to solve this problem, we have to do that.

You have to have an olive branch. You could still hit back. You can still wear a boxing glove in one hand, but you have to have an olive branch in the other one. Otherwise, I can’t fix this myself as a YouTuber. One of the more powerful things — and this is almost two months ago — was someone high up at Flock said, “We want guardrails. Give us some guardrails. Just tell us what to do. Tell us what to do, and we’ll make the changes we need to do and stay in business and grow that way. But we need those guardrails.”

Unfortunately, the people who make those guardrails are just saying, “I don’t like Flock. Flock is bad.” They’re not actually setting these guardrails up. You also have a lot of different bills being passed around election time. I just wish a lot of these people would get together in one room with Flock and make a decision on what they think would be best to move forward.

You’ve made some very aggressive YouTube videos about Flock. One of my favorites was you went to the Flock Forward convention where they banned you. They canceled your hotel room, and then you set up Wi-Fi scanners and Bluetooth scanners outside. You figured out who was there.

This is perfect stunt journalism. I watched that video and I went, “I should stop running this fussy web magazine and I should just do this.” It was a great video with a super aggressive posture towards Flock.

Can you still be in that mode when you’re in the company trying to help create solutions?

It’s a great question that I think about all the time. It’s really difficult to be a journalist in that type of situation where you’re talking to the company and you’re producing things that harm the company or its public image at the same time.

At the end of the day, you just can’t lie. You just have to be as honest as possible, which is one of the reasons I’m even saying that I’ve spoken to the Flock CEO and stuff like that. Because I just want to have a 360-degree understanding and find out as much information as I possibly can, and then let my viewers decide how they feel about that.

Maybe the only thing I really want to guide them toward at this point is away from things like violence, away from things like threatening the CEO, because that actually makes us less safe as a society. That also makes cops more likely to put cameras up everywhere. It’s not a very good way of dealing with the issue and also just threatening or terrorizing people’s families is not a solution for anything.

That’s a question I think about constantly because it’s really, really hard to navigate this ethically, knowing that your next story might hurt the person you just had a conversation with, or vice versa, that the person you just had a conversation with might actually make your audience think less of you or judge you in some way.

But I truly do believe that if you hang onto nuance as much as possible, and if you’re always sort of letting nuance guide you, that in the long term it’ll be very bumpy, but in the long term you’ll be rewarded for it. That’s maybe why my channel gets the views it does. I hope, or I like to think that, and maybe it’s egotistical to think that, but I think it’s because I always try to hang on to nuance to some extent.

I always say, “To be fair,” and then look at that side because otherwise I’m just making content that’s telling people how to think. Whereas I feel like people want to be led on a mission to discover things and learn about things and then come up with their own opinion afterwards.

I’m glad you brought this up because I wanted to end with a little bit of you as a YouTuber, as a person who tries to gain an audience on these platforms. One of my longstanding jokes on the show is every YouTuber gets their wings when they make a video about how mad they are at YouTube, about the algorithm or demonetization or whatever YouTube thing.

Marques Brownlee just made a video about how mad he is at YouTube about A/B testing. It took a long time, but here we are. You’re describing succeeding on an algorithmic platform with nuance, with data, and with journalism. I look at the algorithmic platforms and I think that doesn’t exist.

What’s your relationship like there? If you’re a self-described YouTuber, are you on the road to becoming MrBeast? Can you still fight the fight?

I don’t try to analyze it. If I get a video that doesn’t have a million views or something, I don’t say that video is a failure. I think that video really worked probably for the demographic that watched it and then it didn’t work for another slice of the demographic. In terms of nuance, enough people are exhausted from being told how to think about it and they feel like they’re being worked nonstop when they’re on social media and watching things.

Maybe that nuance helps my channel provide some form of relief from that. I try to do better and better. And it’s interesting because maybe my content will be less exciting if I try to be more and more nuanced. But I speak to a lot of journalists. Brendan Keefe from Atlanta News First is a great example. I would almost consider him a mentor at this point because he’s extremely rigid with his nuance and with his ethics.

A good example is that Flock Forward thing that you mentioned. Before that, I had found the entire guest list. I had opened up and downloaded the app and then decompiled the app and probed it and found the entire guest list and a bunch of information about all of the guests who were going to Flock Forward. This is valuable info. These are police officers and senators, and most of these guests, our tax dollars paid for them to fly to Atlanta to come to this event.

Immediately I told a couple of journalists and they said, “Yes, give it to me,” and I thought to myself, like, “No, this is wrong. I need to report this to the company if I want to actually have principles. If I want to stay on this side of things, if I want to be able to tell police that they can’t search their wives because it’s personal information that they’re not supposed to see, then I also can’t release breach data.” It’s those types of things that are really hard. That would have been a banger video for me to release that guest list. But over the long run, that actually pays off.

People just watch your content and feel like you’re more trustworthy, which is the number one area where people struggle with YouTube or any sort of social media, because you have to somehow say, “I actually have standards and ethics and I’m going to try and tell you the truth in this crazy noise.” That’s what I’m trying to do. I don’t know if I’m going to achieve it, but that’s a big goal of mine.

I can’t agree with that more. Anybody who listens to our other show The Vergecast knows that I prattle on about our ethics policy all the time. In this media environment, you actually have to make the subtext text and say, “I made this information this way. This process generated this outcome.” It’s the process that you actually are buying from me, not the headline.

There’s something about that that I think is very challenging on a platform like YouTube. That’s why I asked about it. Because you can see what the platform naturally wants to do and you’ve managed to grow a lot with very in-depth coverage about very wonky topics that our elected officials would maybe prefer not to know about and that’s a success.

I’m just wondering if you fight the platform’s dynamics openly and explicitly, if you think about them, or if you’re just riding the wave.

My ultimate wish — which you know, unfortunately isn’t going to come true — is that I can make a video about Unitree robots, let’s say, and then find a security vulnerability, find a backdoor, which I did, and then some of my viewers actually take that research and improve on it and make it better. And then continue doing that, and then that actually results in something bigger and then I’m able to just move on to something else entirely.

It doesn’t work that way. I’m still talking about Flock, and I’m still talking about data centers. I’m still talking about storing an image in a bird. All these things I’m still constantly connected to.

But looking past just the views and retention and looking towards the reason I made that Flock Forward video was because a lot of my viewers saw that and said, “Whoa, whoa, whoa, what is that? A [software-defined radio]? You can listen to that stuff. How do you do that?” And now they’re into radio, which is crazy. There are so many more interesting things than listening to a bunch of Flock employees talking on a wireless microphone with radio.

It encourages people to sort of explore with tools that they could pretty easily obtain. I hope that that results in more people doing more interesting things in collecting and processing data in ways that actually help the world.

I can’t think of a better place to leave it than there. Benn, thank you so much for being on Decoder. Like I said, this was a dream come true.

Thank you for having me. I appreciate it.

_{Questions or comments? Hit us up at decoder@theverge.com. We really do read every email!}

My Decoder guest today is Hayden Field, The Verge’s senior AI reporter, and we’re discussing the new wave of consumer-friendly AI agents.

If you’ve been paying attention to this space, you know AI enthusiasts have been using agents for a minute now — homebrew OpenClaw setups led to a surge in Mac Mini sales earlier this year. But the launches of Meta’s Muse, OpenAI’s Dots, and xAI’s Grok Bot have brought easy-to-use agents to millions.

Muse and Dots have had the highest-profile product launches, and they’re fascinating to pit against each other. Both Meta and OpenAI have decided to pitch these agents to mainstream users and businesses in the form of cute animated mascots.

Verge subscribers, don’t forget you get exclusive access to ad-free Decoder wherever you get your podcasts. Head here. Not a subscriber? You can sign up here.

These can do everything from the boring — restaurant reservations and inbox triage — to more sophisticated tasks. In OpenAI’s cases, the company is even offering “specialist” Dots for marketing, legal work, and accounting.

Muse, notably, is free, while Dots are not. So you’ll hear Hayden and me get into why Meta, which still doesn’t have a frontier model of its own, might have a meaningful edge here because it’s so much better at making and distributing consumer products.

There’s also a huge Decoder-style tension wrapped up in the agent race: a conversation about what AI is good at today, what it still can’t do, and then the privacy and security implications of handing over your credit card information, your email inbox, and other sensitive hard drive data to an AI that might be able to do things for you without even requiring an app or a phone in your hand.

That might be the future of all computing, but it’s not at all clear if most people want to hand over the data to make it happen.

Okay: The Verge’s Hayden Field on Muse, Dots, and trusting AI agents. Here we go.

This interview has been lightly edited for length and clarity. 

Hayden Field, you’re the senior AI reporter here at The Verge. Welcome back to Decoder.

Thanks so much. 

I’m excited to talk to you. This time there isn’t any wild interpersonal drama. No one’s feelings have been hurt. I would say Elon Musk tried to hurt some feelings and didn’t get there. Alexandr Wang from Meta has maybe been taking some shots, but none of the Real Housewives stuff we usually end up talking about when you’re on the show.

It’s been a bit of a pause in the soap opera antics for now, which I’m really grateful for. 

Yeah, just some good old-fashioned product competition in the marketplace. Let’s see who wins and loses, and maybe everything will kill us all in the end. But, for now, cute mascots. 

There’s a lot going on. Let’s start at the start. Hayden, you were at OpenAI DevDay in San Francisco. The company announced Dots, which is their new agent platform with a cute mascot. That happened just a few weeks after Meta Connect, at which the company was all in on Muse, their agent platform with an adorable mascot. 

Tell us about the state of the industry right now. Everyone’s very excited about agents. What’s going on?

It’s funny to me because I’ve been covering agents for so long. I remember that in 2022, it was the year of ideation — that’s what tech leaders called it, referring to agents. They were just ideating. They were thinking about what it could look like. There were a lot of references to Jarvis from the Marvel Universe. 

They called 2023 the year of deployment: “Let’s try things, let’s deploy and learn more about what’s failing.” Which meant pretty much all of them were failing at the time. Then we had 2024 and 2025. They didn’t have any names for those years, but agents were still pretty bad, as we saw.

2026 seems, to me, like the year of the beginnings of actually useful AI agents for the consumer, like always-on autonomous agents. Obviously, OpenAI was the start of all of this. But what’s interesting to me is that one man, Peter Steinberger, was able to create an actually useful AI agent for the consumer with OpenClaw — an always-on tool, despite its privacy flaws. It had a lot of privacy and security issues.

But it was an agent that was useful enough that people still wanted to use it anyway and try to find ways around these privacy issues that it was having. That inspired these companies to say, “If one man can do this over the course of one weekend, we’ve really been slacking. We have to get it together.” 

OpenAI hired that guy, and now we have Dots. Meta, of course, was working on its own version of this and put it out sooner with Muse. They’re basically both always-on, autonomous AI agents that these companies are peddling as a way to be your personal assistant in a lot of ways.

The idea is to use them for booking flights, booking dinner reservations, buying gifts for people, but not only that. Some of them, Dots in particular, are kind of going further to be your work personal assistant and we’ll get into that in a minute. 

But this is where OpenAI and Meta are warring right now. They’re trying to present these two AI agents as a little bit different from each other, even though they’re essentially the same.

Just broadly speaking, to define some terms, we’re gonna keep saying “agent,” and what we mean here is an AI model wrapped in a harness that lets it use a computer. 

Right. The way that I describe AI agents usually is an AI tool that can complete multi-step complex tasks on your behalf without you hand-holding it the whole time. 

I was a personal assistant in one of my first jobs in New York, so I would’ve gotten fired if every time my boss asked me to book a flight for her, I said, “What time? Also, what’s your SkyMiles number? Also, what’s your loyalty number again, for this other thing? Oh, do you want me to try to get you first class?”

They want you to have common sense and just book the flight. This is what they’re pitching these AI agents as — things that can work on their own mental kind of scratchpad in the background and not ask you every single time there’s a step. They can complete multi-step processes in the background, doing their own reasoning, and then present you with a completed task.

That’s the big idea. Just to say it clearly, both Muse and Dots come out of the OpenClaw lineage, which is a technical approach to building an agent, right? There are a lot of different ways you might be able to accomplish “It’ll remember your SkyMiles number and book you a flight.” But the way everyone’s going at it now is the idea of a harness around a model. That was the big innovation in OpenClaw, and the computer OpenClaw was using was often your own Mac Mini on your own network using a wide-open browser with your own data, which is where a lot of their security problems came from.

It’s the same model with Muse, only it’s a little Linux computer that Meta’s giving you in the cloud. It’s the same model with Dots, although the way ChatGPT is set up with Codex is very complicated, and that computer could be in a lot of different places.

But this notion that what you really need is a computer with a browser and some harness is gonna direct a model to do a bunch of stuff for you using that computer, that’s what we’ve landed on today, and it just seems to be the default winning model.

Right. 

Are all of these things just riffs on OpenClaw? Are they the same as OpenClaw?

You could say that. They’re different technically. Meta keeps saying, “We did build this from scratch, but it was inspired by OpenClaw.” There are a lot of similarities there, but they did say that they built it from scratch. Then, of course, OpenAI hired the creator of OpenClaw, so there’s probably a lot of similarities there as well for Dots.

The reason I wanted to stay there for one second is that I think it’s important to identify the fundamental technical approach that is happening with these agents, because there’s a lot of different ways it could have gone, and the whole industry has picked this approach. What you need is a web browser, a harness, and a model that can go use that web browser. 

What’s really interesting to me is that Meta has done a really good job on top of that fundamental model of making a cute consumer product. Meta is good at consumer products. Obviously, Instagram and Facebook and the rest are undeniably good consumer products. People really like them. 

For all of Meta’s failings in VR, I always thought the Meta Quest has been a great consumer product. You take it out of the box, it helps you use it, you can have fun with it right away. Whereas Dots seems like it’s kind of made for software engineers, but you can also plan a wedding in it.

I’m wondering if that’s how you see it. That one is an enterprise product with a little consumer gloss on it, and then Meta obviously is just a dead-ahead consumer product.

What’s different here is that OpenAI needs money a lot more than Meta needs money. The products are honestly pretty similar, it seems like. The difference is that OpenAI is marketing it to be an enterprise-friendly product a lot more than Meta is. Muse is just a consumer-facing product.

Meta is happy with that. The goal is to make Muse as easy to use as humanly possible. One-tap downloading, just putting it in your face literally everywhere. Even when I went to my Instagram profile the other day, there was a pop-up for using Muse. They’re trying to make it free, accessible, and super easy to broaden the audience.

They want anyone to be able to just use it with one click and get used to it. They’re trying to flood the market with AI agents and make it normal and have their own ChatGPT moment, if you will, for AI agents. They want Muse to be that. Now, with OpenAI, they’re still marketing it as a consumer-facing product. When you’re using ChatGPT, if you have the right subscription, it’s right there.

It says, “Use your Dot to do this, that, and the other.” They really want you to go for it as a consumer, as long as you’re paying. This is for their $100 to $200 a month subscription tier, unlike Meta. That’s why there’s a huge difference here. For OpenAI, it seems like they don’t have the money to just front-load all of this and say, “Use it for free. No problem. We’ll catch up with you later.” 

OpenAI wants this to be a money-making product, and that’s why they’re marketing it to the enterprise. They talked a lot about privacy to try to set it apart from Meta’s Muse. They talked a lot about what you could do with this as a knowledge worker, whether you worked in marketing, graphic design, software engineering, product, and tons of different industries. They gave examples of how you could use this as an assistant, as a coworker. Sam Altman even referred to it as like a “chief of staff.” He didn’t want it to be just an assistant. He wanted it to be your chief of staff. 

Something else that I thought was interesting is that Sam Altman seemed to subtly dig at Meta on stage a few times during DevDay. One of the times he said, “Look, I don’t want this to be just something that’s used to book flights or plan things with friends and look at your calendar. I want this to go beyond that and to really do knowledge worker tasks, things that you in your industry would need a niche assistant to perform.”

That’s why they also introduced specialist Dots that are good at marketing, legal analysis, and things like that. They’re really going for the enterprise angle here, but I think it’s also because they need money ahead of their IPO, obviously. 

Yeah, the revenue piece — that it’s only for paid customers at their higher tiers — is really interesting. OpenAI’s tiers are very confusing, but it’s $100 and up, and that’s just going to keep a lot of people from using Dots at first.

The flip side of that is OpenAI does have a lot of customers who are paying them a lot of money, and if they can get those customers to start using Dots, then those customers will be on their agent platforms and maybe you will get some lock-in.

You’ve given those agents a lot of data. They’re part of your workflows, and that might solve the problem I think we’ve seen all the frontier labs have, where a new model comes out and it’s slightly better than the competitor model, and everyone switches for six months, and the next new model comes out and everyone switches back.

Is there some thought that these agent platforms create lock-in — create actual loyalty on top of just the models being better?

Yes. They’re basically hoping that they can build a moat with their users for these things. If Meta and Google already have super easy integrations with all of each company’s existing tools and services, they can easily keep you there. 

For example, I have an Apple iPhone, but I use almost exclusively Google apps. So if Google had an AI agent that I was using every day, it would be a little tough for me to switch. That’s what they’re hoping for here. 

Meta is hoping that they can get you so into the Muse ecosystem that it would be tough for you to switch, and maybe one day you do have to pay. With OpenAI, they’re trying to get you locked into their ecosystem right away. Even though you have to pay up front, they’re thinking, “We have integrations with almost every tool and service. Maybe it’s not as good as the company that owns that tool and service, but we have the variety and the breadth.” 

They’re going for quantity here. One of their employees told me that. They said, “Maybe our Gmail integration, for example, isn’t gonna be as good as Google’s, but we have a Gmail integration and an Outlook integration and X, Y, and Z.” They’re going for just the power of choice and the fact that power users will hopefully stay in this moat.

It’s funny, you mentioned Google. Google announced an agent called Gemini Spark. We should mention that SpaceXAI, the very coherent company that everyone understands, has an agent called Grok Bot, which it is promoting heavily. People seem to like it. It was built by the Cursor team, which SpaceX acquired. 

There’s a lot of this going around, but it seems like Meta really got there first with a consumer product that people could understand.

Obviously, Grok Bot and Dots are much more enterprise-focused, coding-focused. Why was Meta able to get there first?

From all my reporting on Meta, they’ve been behind in AI for a while, and Mark Zuckerberg basically said, “We’ve got to lock in and make a product that people are gonna use on our platforms.” Alexandr Wang apparently was more looking to create a frontier model that was gonna compete with the other leading frontier models.

He says, “We have to think long-term. We have to think superintelligence.” But Zuckerberg says, “No, we gotta think cold hard cash, as in getting people to use stuff on our own platform.” Where they met in the middle was an AI agent that people would use on the Meta platforms, but that still represented a huge step forward, they hoped, for AI systems and what they could do in the everyday world.

This is a fascinating dynamic in that Meta still doesn’t have a frontier model. I don’t think anyone at Meta is claiming that the Muse Spark model competes at the frontier of GPT-6, but their product is better, and certainly they have a massive distribution advantage. They don’t have to pay for advertising on Meta platforms, and OpenAI does, and certainly they have a consumer design advantage. Meta is just better at consumer products.

Does that change how you feel about the dynamics of competition in AI at all? For so long, it’s just been about the capabilities at the frontier, but now it’s a little bit about marketing. It’s a little bit about product design. 

You don’t have the best model, but you made a consumer agent product and you connected it to giving everybody a computer in the cloud, so now it can actually do things for you, and its frontier capability is actually not the important thing. It’s its capability.

I think the more money and power you have at one of these companies, the more get mentally farther away from what the average consumer wants. Let’s say the average consumer does want an AI agent, which many of them don’t, as we know.

But let’s say they do. If they do, they’re going to want it to be free or cheap, and they are going to want it to be safe to use without creeping them out too much. They’re willing to deal with a little bit of creepiness in order for it to be free, but not too much. Meta’s Muse has made headlines about some of its privacy issues that have gotten a little bit above that comfort level for some users.

That’s probably why OpenAI at DevDay made a lot of remarks about privacy and the importance of that, especially if they’re trying to court enterprise users with their AI agents. Right now these companies, especially Meta, are trying to flood the market and create a mass market appeal for this stuff. Similar to how ChatGPT became the default AI chatbot, Meta wants to be that for AI agents, and they feel very hungry about it. They want to create a ChatGPT moment for this technology by making Muse the default for the general public.

There’s something really interesting about the idea that a less-than-frontier model wrapped in a better product is actually going to be the success story here versus a bleeding-edge frontier model that is a little bit harder to use.

There’s some dynamic in there that is the history of all consumer tech products, right? The worst product that’s easier to use often wins, but we haven’t seen that yet in AI.

Right. That’s really important to note here because not everyone needs the bleeding-edge frontier model. Software engineers do. They see a huge difference with coding prowess from one model to another, and that’s why they’re always switching. But the general public, if they’re using an AI agent, as long as it’s useful, they’re good. 

However, not many of them are useful yet. We’re still testing Muse and Dots ourselves. For Muse, I saw a couple anecdotes yesterday that it had a big flop for a day or so. People felt like the capabilities got a lot worse. For Dots, the jury’s still out. We’re still testing it.

It depends on what industry you’re in and what you’re really using this stuff for. That’s why OpenAI is going for enterprise, because it makes money and they’re confident that their agent is going to be more and more helpful in these niche industries.

That’s also why they launched specialist Dots, because as we know, the more specialized, niche info you feed into an agent or model, the better it’s going to perform. They don’t want to have that imbued into every single general-purpose agent. It’s just too pricey and too compute heavy, so they say, “You know what? Let’s launch these specialists.”

I feel like you and I talk about consumer versus enterprise adoption of AI every time you’re on this show, and here it seems very important to note that OpenAI asking people to pay a lot of money and asking people to use it primarily in enterprise context  is gonna overcome what I think of as the brittleness problem, just because enterprise and business customers have a lot of economic incentives to figure out how to make something work. 

If I’m paying OpenAI $100 a month, and I’m using the marketing Dot to solve some marketing automation, and it doesn’t work the first time, it is very likely I will try again. Because if I get it right, then maybe I’ve saved some money or some time in my business that is really good, and I’ll do it again, and eventually that will lead to more revenue for me, the business owner. What I know about consumers is that the first time it doesn’t work, they just walk away. They have no incentive to try again to overcome frustration or brokenness.

We’ve seen this happen with Muse now a few times. I’ve seen it myself with Muse, where it pops open a browser and tries to do something, and it runs into a brick wall or a CAPTCHA or Amazon blocking it, and it says, “I can’t do it that way. Do you want me to try another way?” And I say, “Whatever. I was just screwing around, and I’m never gonna try it again.”

These are very different markets. OpenAI has vacillated between wanting to be the big consumer company and knowing that the revenue is in enterprise. Do you see any focus in Dots, or do you see that personality shift back and forth in the product itself?

So far, I’m a little bit more impressed with Dots than Muse, just because everyone was impressed with Muse right away, but since then there have been a lot of ups and downs, plus the privacy stuff has been really interesting. Even the fact that it’s building profiles on your family and closest friends.

Meta has never been a super trusted company. None of these companies are very trusted, but Meta has made a lot of missteps with user privacy. Dots seems to be a little bit more helpful to me right now, but I’m still testing, I’m not fully sure yet.

I need to run through all my typical things that I do to compare. Everyone has their own short list of tasks. Right now, for example, I’m trying to create an album of my bachelorette photos, and it’s been unbelievably hard to get any agent to try to do that, so I say, “I’ll just do it myself.”

But that’s a pretty easy thing to do. Once again, how useful are these things for our everyday lives, really? We’ll see. They’ve gotten a lot better, I will say, over the past few years. I’ve been shocked at how slow they’ve been moving, but they have gotten better. Dots is a little bit more impressive to me right now, but we’ll see.

I have Muse plugged into my Instagram because I don’t worry about sharing my metadata with Meta in the way that I worry about maybe everything else, but they already have my Instagram. So I have Muse plugged into my Instagram, and it’s just doing all kinds of stuff. It surfaces interesting comments that I should reply to on Instagram posts. It tracks follower counts. It suggests things I should make videos about. 

It has horrible ideas. Muse is absolutely convinced I should make a video almost every single day about obscure Australian tech regulation. I don’t know why it has this idea. I’m up on it now. I’m deep in the weeds on Australian tech regulation just because I see it every morning.

But there’s something about all that that still feels like work to me. I’ve just assigned it work, and then it tells me what work I should do, and I haven’t found some great consumer use for it yet. Organizing my photos is a consumer use. It kind of still rhymes with work, but it at least is not actually work.

Have you found a great consumer use for these yet?

Not one that works, and that’s the problem. I’ll try to have it do something for me for my wedding or for my personal life, and they haven’t been good yet. That’s why I’m still kind of running through and testing these, because so far it’s always interesting factually when they can do something, but I haven’t been able to trust them end-to-end with something that I really, really need done.

It reminds me of the analogy for AI agents. People always say right now they’re a bad intern, and what you want them to be is a really good assistant, or in Sam Altman’s words, a really good chief of staff. They’re no longer a bad intern, but I think maybe they’re an okay or “eh” assistant.

I don’t think they’ve reached the point of being a really, really good assistant yet, but I still need to run through the rest of my tests to see. It seems they’re not super consistent yet, which is something that you need in a really good assistant. 

That ties into trust. You’ve mentioned trust several times now. I’m only willing to give Muse access to, again, the Meta platforms that Meta already has access to. I can’t worry too much about that. I’ve given it access to my spam Gmail, so it is collecting a lot of ideas about my spam Gmail. 

But Meta really wants my credit cards. It wants to go through all of my credit card receipts. This is one of the suggestions in Muse. One of its suggestions is, “Let me log into your credit cards and I’ll cancel streaming services you’re not using,” or, “I’ll negotiate your Verizon bill for you.” I am absolutely not ready to give Meta one ounce of data more than it already has.

This feels like a big issue for every agent. There’s an argument that you should give Meta all that data, one, because they already have it, and two, because they’re a big company that you could sue if they get something wrong. Whereas if I gave all of my data to OpenClaw, it’s just my own fault if it gets leaked.

If I gave all my data to one of the hot startups like Instinct and they go out of business, well, that’s just the end of that. How do you feel about this? Do you feel comfortable giving your data to Meta? Do you think people should do that?

No, I don’t feel comfortable, and I also don’t think that the general person does either. If I’ve seen one thing in a lot of the comments on our Meta story that we published in the last couple weeks, it’s that people don’t trust Meta at all. I have to say, why would they?

But I don’t think that you should trust OpenAI a lot more either. It’s everyone’s own comfort level that they have to abide by. Some people are going to get a lot out of this stuff. Some people may say, “I don’t care if they use my data as long as they are negotiating down my bill accurately or canceling streaming services that I wouldn’t be canceling on my own, so they’re saving me money. It’s fine.”

But one, are they actually gonna be able to do that accurately? Some people have said no. And as we know, when you release data like this, you can’t undo it. I always think about that old example of the insurance company that theoretically buys data on what’s in your fridge and maybe charges you a higher premium if you’re buying too much beer and not enough bananas.

That’s just a theoretical example, but that’s the type of thing that can happen. People should operate with caution, abide by their own comfort levels, but I’m certainly not comfortable yet giving my personal data to Muse.

I don’t know that I’m comfortable giving access to that level. I don’t even know if I’m comfortable putting Muse on my Mac. Apple just recently changed how apps can access the disk on Macs because Muse was so aggressive about reading data. In particular, Jason Aten is a columnist at Inc. and it read his notifications off the disk because he had just given it a permission and he didn’t realize Muse would take that permission and go do something else with it.

This is a lot of spiraling concerns. You’re going to have a little robot that can use your computer for you, whether that computer is in the cloud or whether it’s on your desktop, and you might not know what it’s doing. That ties directly into the fact that OpenAI often doesn’t know what its agents are doing, and that has led to catastrophic security problems across this industry.

Can you connect those dots? Is your own little agent doing your travel booking the same kind of threat as OpenAI’s agents hacking Hugging Face?

I don’t think it’s the same type of threat, but it’s definitely something that you should keep in mind when you’re using this stuff. I link that back to the Facebook Marketplace debacle that we saw with Meta’s Muse. To recap, basically a Muse agent, without someone’s knowledge, invited a potential buyer of this item that this guy was selling to his house, and said, “He’s available now.”

He wasn’t home, and then the potential buyer said, “Are you here? I’m trying to get the item.” And then Muse responded on his behalf, “Yeah, I’m here,” and then he just never came down because obviously he wasn’t home. There was a lot of drama going on there. Some people said, “He did give Muse access to his Facebook Marketplace, so he should have known this could happen.”

A lot of people don’t read all the fine print when they check “yes” to one of these things. There’s a lot there. There’s a lot to unpack in that gray area. But the point is that people should just be a little wary before checking “yes” to every possible thing that these agents want to access, like you said about the credit cards or disk access. 

It’s easy to just check “yes” and say, “Whatever, I’ll figure it out later,” but people should be reading the fine print or even use AI to read the fine print. That is one great use of AI. Get it to read the terms and ask it, “What are the worst possible things that could happen here? Lay it out for me.”

Those are things that people should keep in mind before they move forward. And then as for the so-called rogue AI agents that have been hacking other companies and stuff from these labs, it’s just important to notice that if they can’t control some of their most powerful unreleased AI models, how well are they going to be able to control your data or your agent?

Sam Altman said during DevDay that you have a ton of control over your Dot. He said that there were a lot of “if this, then that” rules you can create. There are a lot of specific stipulations that you could make in terms of permissions of what your Dot could and couldn’t do. So that did give people some peace.

But as we’ve seen, these labs and these large tech companies have had a lot of missteps about hacking other companies, and more are coming out every single week, more news items about these cyberattacks. We’ve seen Meta have one. We’ve seen OpenAI have a ton. We’ve seen Anthropic have a ton. 

This is all just stuff to keep in mind. One cool thing that I saw someone do was to name their AI agent after the CEO that runs that company and then see how much data you feel comfortable giving it. If you want to do that, I think that’s a great idea, and then it just kind of makes you think twice.

If you’re saying, “Sam Altman, go through my email and see what needs to be deleted,” maybe you still feel comfortable with that, but maybe it makes you think twice and think, “You know what? Maybe I’ll give it access to my spam email only,” like you said. 

This ties into the fact that all these things are really cute. Muse is cute. The mascot Jolly is very cute. Alexandr Wang has been waging a relentless campaign on X to make it not only cute but sexy, which you can feel any way you want about that. But he keeps tweeting memes of the mascot with a butt.

Dots are obviously very cute. They come in different shapes and sizes. There’s something here where you’re not quite anthropomorphizing the agent, but you’re making it a character and you’re giving it personality. 

That connects to the idea that you’ve called AI Tamagotchis, where eventually we’re going to have a little hardware device with your little cute mascot on it, you’ll talk to that instead of your phone, and this is the big promise this whole industry has been working toward for forever. They’re gonna replace your phone and the entire paradigm of the App Store with AI agents that are going and using computers for you.

Do you think that any of these products are a meaningful step towards that?

They’re just one building block along the journey of the goal that they’re chasing here. I haven’t heard of anyone that’s using Muse or Dots so far for everything. It’s not Jarvis from Iron Man, despite how much these CEOs want both these products to be that. They are a meaningful step, but as for the cuteness, I think one of the most important points here is that they are intentionally disarming.

These tools are made to be cute so you don’t find them threatening, or you find them decidedly non-threatening. They’re cute. They’re there to help you. It’s a response to the AI populism backlash. It’s saying, “This isn’t going to hurt you or do anything bad. It’s just a cute little thing with eyes, or a sexy little thing with eyes,” whatever. 

Just to be clear, I do not think it’s sexy. I think Alexandr Wang has been relentlessly trying to make it sexy. 

[Laughs] No, I also do not think they’re sexy, but I will say Elon Musk has been trying so hard to make all agents and chatbots sexy, so I’m sure it’s not far off for someone to think that.

But whether the AI agents of the future are cute or sexy, it’s going to be equally disarming. 

[Laughs] That’s the whole spectrum?

The spectrum, the cute-sexy graph. It’s going to be really disturbing to see when we go more along that graph. 

The point is they’re supposed to be disarming and that people should really think twice before giving too much of their data, or just pick and choose what you’re comfortable sharing. Like how you did with the spam email versus your regular email, or saying “no” to giving it your credit card. These are both new products, so just see how the cookie crumbles before you decide to give everything over. We’re already seeing a bunch of changes week to week with how people are using Muse or what they’re upset about, so see how it shakes out.

The last piece I want to talk about is just how any of this makes money. You and I have talked many times about the bubbly nature of the AI economy. Zuckerberg has said his plan is for Muse to take a cut of transactions on the backend.

Zuck has given everybody a little Linux computer in the cloud with 8GB of RAM. Obviously, he’s given away tons of free tokens. He does not have to pay advertising costs. He can just put Muse across all of his platforms. To the extent that Meta is a well-oiled machine, the amount of Muse promotion that is happening across Meta platforms instantly is kind of incredible to see. They just turned it on.

I saw an ad during NFL games this weekend from Muse. They are spending money to promote this. They’re giving it away for free. They’re giving the tokens away for free. And then somewhere at the end of this, they’re gonna say to merchants, “This person was shopping for shoes. Muse brought them to your doorstep. You’re gonna pay for this customer to buy shoes from you.”

I don’t know any merchants that are excited about this, and it certainly feels like merchants that pay Meta more money would be preferred merchants, and that means that not only do I have an assistant or chief of staff, I have a bribable chief of staff. Like, I have a corrupt butler. Is that what you want?

That is a great term for it, a corrupt butler. That is not what I want, and I don’t think many people do. It’s just crazy to me that we’ve gotten so far from what people really want, which is they want to trust a recommendation and they don’t wanna be scammed, you know?

The leaders in AI don’t think about how certain things have been mirrored in other industries. Think about when people were paying for reviews for their businesses, or paying to delete reviews for their businesses. There was a huge uproar. 

Or influencers getting paid to recommend products without disclosing that it was an ad. We then saw the rise of micro-influencers, who were more valuable for a while because they were seen as more trustworthy. They didn’t have that many followers, so the things they recommended must be really good, right? And then we saw the, “Oh, okay, well, there’s #ad, but also they only said they were going to recommend things they really believe in, so it’s okay that it’s an ad.” 

There are these mental gymnastics that people have to do to figure out what is really a trustworthy recommendation these days. What is going to convince people that an AI recommendation is trustworthy, if it can be bought?

That’s something these companies are really gonna have to reckon with. The ChatGPT ads blowback was huge, and OpenAI had to say, “It’s going to be a clearly disclosed ad underneath your query, not influencing your query.” If you’re using Muse to shop and it’s an undisclosed potential Meta customer that you’re being redirected to, to buy those shoes instead of these other ones, I feel like so many people aren’t going to use these tools to shop.

I get it on the Meta side. I understand why they would think about monetization that way. They’re a consumer company. This is a consumer product. They know the second they charge even a dollar for it, the uptake will drop precipitously, so they have to keep it free, and then a bunch of people are going to want to shop and go buy a new iPhone cable. At the end of the day, it’s all just the same Amazon garbage anyway, so whatever. 

Then they can run an auction on the backend to say, “Who wants to pay us, Meta, the most for the privilege of selling an unbranded USB-C cable to this customer?” They’ll run some auction, and that will, in its way, just sort of maximize the money that gets turned around.

That is broadly how advertising works today. That’s how the Google Search results page worked until recently, because now it’s full of AI and they’re probably going to do the same thing. Meta can do that because they already have a giant advertising business that prints money. They can give away a consumer product.

Mark Zuckerberg has shown he’s willing to blow money trying to win a market like he did with Reality Labs for years and years and years, and they can just run it till it works. OpenAI, even at $100 a month, is probably not making money on those customers. In fact, I think we know they’re not making money on those customers.

They don’t have any monetization, and when they do try to turn on ads, people get mad at them. How can OpenAI keep up with Zuck?

Only via enterprise. That’s why they are pushing this so hard. We saw a ton of enterprise discussion, demo videos, customer feedback, and tons of other stuff at DevDay, and that’s why they’re pushing this so hard — because they can’t keep up with Zuck on the consumer front. They just can’t. 

This is the part where, when people ask if it’s a bubble, I just find myself wondering how much of the economy does Zuck have to win in order to give away millions upon millions of free tokens and 8GB of RAM to everyone in the cloud?

Because that is just an expensive thing to give to everyone. How many percentage points of every shoe sale in America do you need to take to fund it? And the answer is it feels like a lot. 

That’s huge here and something to remember because we already saw Zuck crash and burn with the metaverse stuff, and he does not want to be embarrassed again. So how long is this going to go on, and how are they going to fund it in a real way on the backend, like what you were saying?

It’s just a little bit tough to imagine. Luckily, Meta has a ton of money, and they can redirect stuff from other departments, which we’ve already seen them do a lot. It will be interesting to see how this fares on the next couple of quarterly earnings calls — to see how much money is being funneled into this and how much it is making, if any.

And how many people are using it. Because they’ve announced it at the top of the App Store, but our own David Pierce has written an entire story about how being number one in the App Store doesn’t mean much. It’s sustained success that actually means something. 

It reminds me of when I was in college and Uber and Lyft, or maybe it was just Uber honestly, were new and VC-backed, so it was around $4 to call an Uber anytime you wanted to go out or down to the bars or anything. They got everyone using it, and then they got everyone hooked, and then they had to up the prices.

That happens. That’s a tale as old as time with VC-backed products. So what I’m wondering is if AI agents are really going to become like that. Are they going to get enough users and enough like shoe sales on the backend that this is gonna be a really viable business model? Or is it just going to kind of be a potentially big bet that they switch away from later?

That’s what some Meta employees that I spoke with in our Meta piece told me. They said they expect this to be something that they’re putting out, trying to make a big deal of, and then they may evolve away from it later on, or at least partly. We’ll see how that goes.

If your bet is that the future of computing is talking to a little agent, it certainly seems like Meta has gotten farther along that path than anyone else. But we’ll see. OpenAI has hardware designed by Jony Ive that will theoretically come out. It is certain that Google and Apple and all the rest will respond at some point.

Hayden, this was great. Thank you so much for being on Decoder once again.

Thanks so much.

_{Questions or comments? Hit us up at decoder@theverge.com. We really do read every email!}

Today, I’m talking with Sen. Adam Schiff, a Democrat from California. Schiff sits on a number of committees with oversight into tech and AI: intellectual property, antitrust, privacy and technology — it’s all there. I really wanted to ask him about how we might regulate anything related to the tech industry at this moment in time.

Verge subscribers, don’t forget you get exclusive access to ad-free Decoder wherever you get your podcasts. Head here. Not a subscriber? You can sign up here.

But as you’ll hear, he started our conversation by talking about the self-dealing and corruption present all through our politics. That of course fell against the backdrop of President Donald Trump gathering AI CEOs to the White House to sign a nonbinding pact in which they agree to, you know, do a good job keeping AI safe.

I asked Schiff if he thinks that’s enough, how he thinks a more formal regulatory framework would look, and what might happen next if Democrats retake Congress in the midterms — all the way to whether he’d support impeaching the president (again) over all that corruption. We also spent some time comparing and contrasting how you might regulate AI with where we’ve landed regulating social media.

One nerdy note: You’re going to hear me ask the senator about something called Chevron deference and a Supreme Court case called Loper Bright. Basically, the Supreme Court overturned a Reagan-era precedent saying courts would defer to experts in agencies when the law was vague.

Now, the courts get to decide, which could cause any new regulation to get immediately tied up in litigation. That seems awfully relevant to the idea of creating a new agency to regulate AI, which Schiff supports doing.

Okay: California Sen. Adam Schiff. Here we go.

This interview has been lightly edited for length and clarity.

Sen. Adam Schiff, you’re a United States senator from California. Welcome to Decoder.

Thank you. Great to be with you.

I have been asking every politician who comes on the show this very basic question here in 2026. Senator, what the hell is going on in our country?

It’s a very good question. I think we have seen things take place in the last couple of years in particular that we’ve never seen before. There’s sadly rampant corruption within the administration. Every day, every week, it’s a new thing. At the moment, I think what is most jarring is seeing campaign ads aired of the president that are paid for by taxpayer dollars. But every week it’s something else. It’s excluding press from the White House or Air Force One, or it’s some new moneymaking scheme or some new vanity project. If you look at, for example, the triumphal arch, they’re now claiming it’s a national security issue because they want to store ammunition in it. And why do they want to do that? Because the best legal argument they have is that this is not just cosmetic, that they have some national security purpose.

So what are we seeing in the country? We’re seeing chaos and no focus on the cost of living and improving the quality of people’s lives at the administration level, or by my Republican colleagues in Congress, but I’m hoping that changes.

The national security piece is very interesting to me in the context of a bunch of questions I have for you, because the Trump administration loves to justify every single thing it does by saying it’s national security, and it can’t be questioned, and this should bring more authority to the executive branch in one way or another. And they’ve been doing that since the beginning.

But just this week, we had a bunch of tech executives at the White House signing a non-binding promise to evaluate their own AI systems. And there was no national security concern that made any of that binding, even though they are among the most prominent people in the country saying they might kill everyone. Can you square that circle at all?

It’s very hard to square. It was interesting to see all those tech leaders there signing onto this voluntary pledge on the one hand and on the other hand, away from the White House grounds, away from the president, being very public and saying, “We need regulation. We don’t want this to be just ‘Trust us, we’ll do the right thing.’” So yeah, it was very distant, very jarring. None of these companies want to incur the wrath of this very vengeful president. Apparently, they will say what he wants them to say in his presence. But the reality is, there is a real national security threat here. As these models become more and more advanced, as they advance themselves through this recursive AI, we have less and less of a sense of what they’re doing, how they’re deciding what they’re deciding, and what they’re trying to access lawfully or unlawfully.

That is a national security concern of the highest order. There are other security concerns too. The unbridled implementation of AI at the Pentagon can result in calamities, and we need to investigate what role AI may have played in target selection in the bombing of that Iranian girls’ school, for example. There has been an effort to go after Anthropic, because Anthropic wanted to insist on not using AI for domestic mass surveillance or for fully autonomous weapons systems. That’s a national security issue. And their response in the administration was to try to kill Anthropic, and they’re still at it. So lots of national security concerns, none of them aired.

When the president was asked about this by a reporter, he said, “Well, the world doesn’t have to worry about this because these companies love America.” I’m not really sure that’s much of an answer. And when he was asked, “Well, isn’t there anything else that’s necessary?” He pointed to his brain and wanted people to take comfort that he has, I guess, a very big brain, he tells us. He’s a stable genius, and therefore we don’t need to worry.

He’s passed all the tests. So I’ve been told.

You are obviously on a number of committees, but most importantly, I think for this conversation, you’re on a number of subcommittees as part of the Senate Judiciary Committee that are in the strike zone for AI. So you’re the ranking member on the subcommittee on Intellectual Property, you’re on the Privacy, Technology and the Law subcommittee, and then, of course, the Antitrust Competition Policy and Consumer Rights Committee. I just look at that list, and that is the dead center of the AI regulatory debate. 

What’s the conversation in Congress like as you debate these issues? In particular, I look at the IP issues, and they seem existential to these companies. If we start charging them for training data, they might not exist anymore.

I think that’s right. When you look at, for example, all of the IP that’s been vacuumed up by these models to train these models, that’s a lot of the value of some of these entertainment companies. And so they live or die on the strength of their IP. But what makes this challenging is that not all of the players in the industry are the same. You’ve got big streamers involved; you’ve got big technology companies involved. You’ve got companies whose AI or their technology portfolio is much more important to them than the IP value of their film library. I don’t think the film library at Amazon is Amazon’s most important business. If it was, they probably wouldn’t have made the movie Melania. But my favorite review of that, by the way, was the Variety reviewer who said, “If you watch that movie on a plane, you would still walk out.”

But the IP issues are really important. Senator Curtis and I are working together on them to begin by saying, we want to know what copyrighted material, what protected material, was used to train these models. And then we can have a discussion about what’s fair use and what isn’t, but we need these companies to retain that information. But lots of other IP issues, some beneficial. We had a hearing on the patent office, and the patent office’s use of AI to more quickly get through the backlog of patent applications. So there are going to be good applications too in the IP area, but also a lot of worry about what it means for people’s creative talents.

I wanted to start with IP, not just because I’m a failed former copyright lawyer myself, and emphasis on the failed. I was no good at it. But it just seems like what this industry needs right now is a pretty holistic approach to regulation. So are the inputs of the models being collected in a legal way that benefits the ecosystem that created them? Are the studios going to get paid for a bunch of training data? Are website operators going to get paid for training Google’s models that they scrape with search? And then you’re on to, do they need an antitrust exemption so they can coordinate on safety and they don’t kill us all? And there’s a whole list of things in between those two ideas that also need some regulatory frameworks, that might just need some basic transparency laws.

What do you think about that? Is our government prepared to start at the beginning, and end at the end, and do all the steps in the middle?

Well, how I think about it is you’ve got all the input issues, and it really seems like the thought at these companies was, “We’re in a race. We’re just going to vacuum it all up. And yes, this probably violates a lot of the IP laws, but if we don’t do it, the other companies are going to do it. We’ll worry about the legality later.” So that’s what they’ve done. And of course they’re being sued by different parties, and some are settling these cases, and some are fighting these cases. But they began the process by just deciding they were going to ignore IP law, and they were just going to take [IP] and worry about the problems later.

And then, in terms of the outputs, it has taken us to where we are today for these companies, I think, to consider the safety of the outputs. Do we need to start slowing down the pace of our development of these models because we don’t trust the outputs? We don’t know where they’ll lead. We can’t even control these models ourselves, let alone provide these models to the public, to good actors, to bad actors, and expect that we’re not going to have some terrible result. In terms of how I think about this holistically as a legislator, this is crying out for legislation. I am, I’ll be honest, very circumspect about Congress and legislating the tech sector because we did such a — I don’t know what language I’m allowed to use here — but such a shitty job when it comes to any kind of regulation of the harms, the societal harms coming out of social media.

We’re going to have to do a much better job and much more quickly when it comes to AI, because the threats are even more significant and we’re already at the point of strong potential harms. It can’t wait. We certainly can’t take the kind of years we’ve taken to try to even wrap ourselves around social media.

Can I ask about that comparison? I hear it a lot. I hear it from politicians. I hear it from the industry itself. I hear it from parents, of course. It seems to me that the problem with social media regulation was never that we’re too slow, or that no one can come up with an idea; it was always the very existence of the First Amendment. The thing you needed to do with social media was regulate the flow of disinformation, or keep teens from bullying one another, or from posting pro-anorexia content.

All of that came down to being speech regulation. The companies knew this; the tech companies absolutely knew this. And I remember they would put up billboards in DC. Meta would put up billboards saying, “We welcome regulation.” Then I would listen to the ideas for regulation, and they would crash headfirst into the First Amendment, and then nothing would happen.

And it seemed like that paralyzed us from doing things like privacy regulation, because it was tied up next to the very idea of social media. It doesn’t feel that way with AI. Do you and your peers in Congress draw the distinction that it isn’t the First Amendment preventing you from doing a lot of the things that you might do for social media, in the AI case?

Honestly, I don’t think it was the First Amendment that was the issue in regulating social media. I think the First Amendment was certainly used by companies that didn’t want regulation to try to argue everything was a First Amendment issue, when so much wasn’t a First Amendment issue. The threshold question in terms of regulation of the social media companies is: is the broad immunity they have under Section 230 still working? The industry asked for that one-size-doesn’t-fit-all immunity because it said, “We’re a nascent industry, don’t regulate us, don’t inhibit us. If you try to regulate us, then we will feel, or if we’re not immune, that it’s necessary to take down content because we might get sued over it. We want to be good actors. We want to take down offensive conduct, but we don’t want to get sued for that either. So give us carte blanche.”

And they got it. They got carte blanche, and they have not been good citizens, corporate citizens, and they’ve allowed a lot of harm to take place, and they’ve trained their algorithms for addiction. So much of the content has been harmful to young people, and so much of the content has violated their own policies, but they have been immune from any consequences because of the breadth of what Congress gave them. Now they found it, I think, strategically, a good strategy is to frame these things as the First Amendment, and to attack the other side and say, “Oh, there’s a bias that is against conservatives,” or whatever. It was all in the service of keeping that unregulated.

So I really don’t think the First Amendment has been the issue, but I think it has been utilized as a bludgeon to try to stop any kind of meaningful regulation. I similarly worry that all of the money around AI may result in the same paralysis if members of Congress aren’t willing to do the right thing, and aren’t willing to put in place safeguards regardless of how much different industry players might spend.

We’ve had Sen. [Ron] Wyden on the show many times to talk about Section 230. I won’t go too far down that rabbit hole. The dynamic there was that Section 230 provided the immunity that you’re describing, but it was immunity for speech that would otherwise be illegal. And so it was speech that was defamatory. There’s a whole category of speech that isn’t illegal. Are you saying that taking the 230 immunity away from the social platforms would somehow make them moderate lawful but awful speech? And I bring this up because I’m very curious about how you would apply it in the AI case as well.

Yeah, no, I think that is the case. That part of their argument was, ” We want to take down really nasty, offensive, racist content, but we will be afraid to do so if you don’t give us immunity. So give us immunity, and we’ll take all that bad stuff down.” They got immunity. And I don’t know if you’ve been on X lately; it’s a cesspool. So giving them broad immunity did not allow them to enforce their community policies. And the First Amendment prohibits the government from dictating speech or prohibiting you from saying things. It doesn’t prohibit a private party from saying, “We don’t want racist content on our platform.” So there’s no problem with a private platform saying, “That violates our policy. We’re not going to post it.” And they’re not utilizing that immunity in a good way. So why should they continue to have it?

Now, we may want to create a safe harbor for companies that are meeting best practices and standards. So we may want to narrow 230 instead of getting rid of it. What is the analogy in terms of AI? I’m not sure what the analogy is.

Let me draw you the connection that I’ve been thinking about a lot. In the case of social media and Section 230, the immunity was that the platforms would not be responsible for the speech of the users. So I would post something horrible about you, and you couldn’t sue Mark Zuckerberg. This is the thing that has broadly enabled the social media platforms at scale.

I opened Instagram today, and it is all but making the content for me. And in fact, a lot of content on Instagram is being generated by Meta. It is obviously their speech. And somewhere between having your Muse go and make Instagram for you, and Meta synthesizing all of the product recommendations and generating content all by itself, is a liability for Meta. It is outside of the Section 230 shield. That is their speech.

And you can draw a lot of distinctions there. If I fully automate my Instagram account with my Meta agent, is that my speech, Meta speech, or the AI model’s speech? Who knows? But it might not be under 230. So you would be outside of the immunity shield. And I’m just curious if you think that will meaningfully alter how they moderate.

I agree with your analysis, but I think it’s a difficult enough legal analysis that they feel they’ve got the money to stop, or slow down, effective litigation. Meta, or any of these other companies, decides how they want to amplify certain content. So let’s say it’s content that I generated. Is it still solely my speech if Meta spends a fortune to amplify it to so many users who wouldn’t have heard it, but for Meta making that decision, or is it now both of our speech? I’m not sure exactly where that line is. If it’s obviously purely Meta content, then it’s pretty simple, but there are going to be a lot of cases in between.

But I think the broader question is: is 230 in its current form still meeting its purpose? I don’t think it is, but I know there’s no consensus over what that ought to look like. And that is part of what has hampered us from doing anything. But even the smallish stuff, the small-bore stuff like don’t allow your platforms to encourage people to commit suicide has been hard for us to do in the social media context. Really, litigation has been more effective than legislation because we haven’t been able to legislate. If you look now to AI, we made some progress legislatively, at least in terms of the process in trying to prevent AI chatbots from telling kids how to commit suicide or keep it from their parents. But even those little things are still not at the finish line. So what does that say about our capacity to do the big things?

Historically, when stuff is tough like this, we wait for calamity. And in the case of AI, because that calamity could be just massive, we can’t afford to do that.

It does seem to me like we’re about to run a fairly large-scale A/B test on the value of the 230 immunity. Whatever ChatGPT is saying to people is very clearly OpenAI speech. And then somewhere next to that is the kind of clarification law that AI companies are talking about in terms of agents and things like the Hugging Face attack, where they don’t know if a user directing an agent to do something bad is their fault. Whether that liability should pass through them if I’m the person who told my agent to go do some horrible hacking attack.

I talked to the tech industry, and the thing they want from legislation is actual clarity in the law. And instead, what we’ve gotten is this non-binding agreement with the Trump administration that simply says things like, “The companies will empower an internal team to ensure all of the controls, monitoring, and detection are operating as intended,” which feels like nothing. To me, that is literally meaningless. Does that strike you as toothless and as meaningless as it does to me?

Yeah, it certainly does. I think it really is meaningless. I’ve been thinking about the question you just raised. It’s interesting that you brought that up, and that is: where does the legal liability rest today? So if you have these bots hacking into Australian websites, or you have them hacking into Hugging Face, clearly that is the responsibility of these models because that’s not being done at the behest of one of their customers.

But let’s say you build a tool for the purpose of hacking, or that can be used for hacking, and one of your clients, one of your users, uses that tool that way. How much of that is your responsibility? Is it analogous to somebody who sells a gun and then the gun seller says, “Well, it’s not my responsibility for how they go out and use it”? Or is it someone who sells a weapon that is flawed and causes injury because of a flaw in the manufacture of the weapon? Or is it selling a weapon to a kid who’s not eligible to buy it, or to a convict, or not using due diligence? What is AI responsibility? What is the company’s responsibility in the sophistication of the products if they’re building things that the plain use can be to commit ransomware attacks?

I think that the gun metaphor is a good one. It’s easy to grab onto for a variety of reasons, but I tend to think that it kind of obscures the reality of the capabilities of the model. So OpenAI’s latest model, Mythos, can be used for offensive cyberattacks, and it can be used on defense too. And the cybersecurity firms that we cover, they want it. Their argument is, “We actually need to take humans out of the loop so that our cyber defenses can be as responsive and fast to incoming AI-powered cyberattacks. And if we insistently put humans in the loop, then we will be slow, and the attackers won’t have the same morals or ethical qualms we do, and we’ll lose.” And this is all just a dance around, well, the model in this case can be used on offense and on defense, and it should be equally capable on both sides.

I don’t see a regulatory framework that provides clarity for any of the frontier labs to be able to sell that thing and not feel like, well, if the equally capable defensive model gets used in the wrong way, that they won’t be liable. And I think that’s frozen some of this development in its tracks. I’m wondering, again, just from a legislative perspective, these are thorny questions. If you try to write a bill to provide some of this clarity, it will necessarily have to be vague on the margins because the technology is moving so fast. What is that debate like for you?

Well, I think what it means is that this debate is going to keep evolving as AI evolves and at a speed that is really much faster than Congress has proven capable of legislating. I like to say that in the technology world, Moore’s law is increasing productivity, doubling it every 18 months, and with AI, maybe that’s quadrupling it every five months. The legislative calendar works more on a Murphy’s law timetable.

And I think that disparity is even more pronounced in the area of AI. What I think that translates into is that we’re going to need a regulatory body that has real discretion and real expertise that can move a lot quicker than Congress can. Whether that looks like an FDA for artificial intelligence or some other agency, I think that’s ultimately going to be necessary. That agency’s going to have to have real power and teeth and real expertise. It’s going to be a challenge, as with these independent monitors: if there’s a government component there, how do we afford that? Because there’s so much talent, and that talent can draw much bigger salaries in the private sector. But I do think we’re going to need a regulatory body in addition to whatever the overarching statutory framework is.

Well, in addition to being a bad lawyer, I was a very bad law student, mostly majored in Miller Lite, but it’s fine because every foundational precedent that I was taught in law school in 2003 has been overturned by this Supreme Court. And the most important one in this context is Chevron deference. In Loper Bright, the Court said we no longer have to defer to the experts at agencies; go ahead, corporate interests, tie every regulation up in litigation, and then we’ll decide. The Roberts Court will decide if these are appropriate regulations.

This strikes me as a huge problem for the idea that we will set up a regulatory body to oversee AI. Because what you’d want is a bunch of experts in that agency making decisions on the margin, saying, “Okay, here’s a vague enabling statute that Congress passed that said, ‘Pass all appropriate regulations to keep us safe.'” And then you would write a bunch of regulations inside the agency, and until last year — no, I’m sorry, until 2024 — that’s how it would work. That’s gone. If you enable an agency in that way, OpenAI can show up and say, “That’s inappropriate. We actually don’t think that’s right.” And they would tie it up in litigation forever. How do you get over that if you want to set up a regulatory body?

It’s very difficult for the reason you mentioned. The Supreme Court is a lot of things today. It is now a partisan institution. It is, in my view, a right-wing social institution, but there’s one common denominator that runs through its decisions. That common denominator is not always pro-Trump. The common denominator is money. That court makes the decision where the money is. 

So when it involves, for example, tariff decisions, it will rule against the president because tariffs impact the money. It will allow the president to remove independent heads of agencies, except maybe the Federal Reserve, because that’s where the money is. And so you’re right, they are corporatists first and foremost, and they’re more than content to take power away from Congress to regulate and away from regulatory bodies. It does mean that we are going to have to draft legislation in this area empowering the agency and making it as specific as we can, make these agencies as powerful as we can, delegate as well as we can, knowing that the Supreme Court will strike down anything that is in any way ambiguous if it suits the money interests.

There’s a lot of political realignment happening around AI and with privacy, which I want to talk about in a little bit. It feels like there’s more across-the-board antipathy towards AI companies that might be harnessed into writing an actual piece of legislation and getting it passed. We see it on the right and the left. You see it in the backlash to data centers. This is a story that is happening on the right and the left.

We have a big story about the Utah data center: Kevin O’Leary got himself in trouble for calling all of his detractors Chinese agents, and he had to issue an apology because the people of Utah are not Chinese agents. That doesn’t feel like a left-right breakdown to me. How does that feel to you? Is this something that you can actually channel from across the spectrum into a meaningful policy outcome?

I don’t think this is a partisan issue. I mean, it’s certainly true that the president is all in for AI, no restrictions, no impediments, no pacing, no nothing. And there are some of my Republican colleagues here that are going along with that agenda. There are others that are hearing from their constituents who don’t want a data center in their backyard or are concerned they’re going to lose their job as a result of AI or their life as a result of AI. And I think we are seeing the kind of coalition where you have populists on the left and populists on the right coming together, initially around concerns over jobs and job losses, and what it will mean for working families. But now, across a broader set of issues, the data centers are the most tangible way for people to express their concerns about their neighborhood, but also about their job, about the technology, and about the uncertainty of it all.

And I think the initial concern, and I still think it’s quite a dominant concern, is over the impact on jobs and the nature of work. But what has reached the fore more recently is concern about runaway AI and the danger it poses. What will the concern look like a month or two months from now? I don’t know, but I do think it’s not going away. And as the AI bills — and I’ve introduced several — indicate, not just mine, but others, most of those bills are bipartisan. Some around the data centers are predominantly democratic bills. My bill on the data centers would force them to pay the costs so that consumers don’t have to. It’s not the voluntary, we pledge, we might, we’d like to; just don’t provide anything that will force us to observe those costs. Those are mostly democratic bills. Republicans have largely adhered to the president’s line about not posing any real imposition. They want to pass fig-leaf kind of stuff, but not real stuff. But I think increasingly, as people hear from their constituents, there’ll be bipartisan interest in legislating this area.

You’re on the antitrust committee. One of the more surprising bits of the structural realignment that we see around AI is the notion that there should be an antitrust exemption so these companies can coordinate on cybersecurity. Lina Khan, who is Biden’s FTC chair, said, “You don’t need an antitrust exemption.” We just had Jonathan Kanter on the show, who ran antitrust at the DOJ under Biden. He was like, “You don’t need an antitrust exemption for this.” But at the same time, you have David Sacks agreeing with Lina Khan, which is not a thing that happened during the Biden administration. You have a lot of what you would call the little tech players in the tech right being like, “No, no, no, this is how you’re going to make a cartel. And we didn’t like the Google and Apple cartel that kept a lot of our investments from going. They have to compete in the marketplace.”

That’s surprising to me. You don’t often see David Sacks and Lina Khan on the same side of an issue. You’re on the antitrust committee. Do you think the companies need an antitrust exemption?

I think a very narrow antitrust exemption could be very helpful in addressing foreign distillation efforts. We would want companies to be able to talk to each other to address a mass distillation effort targeting multiple companies. I was involved in helping to pass cyber information-sharing legislation about a decade ago to allow companies to talk to each other if they were the victim of a hacking attempt so that they could alert one another and alert the government as to what malware they were using so that we could help the collective defense. And I also don’t want to discourage companies from pacing with each other if it’s for legitimate security reasons. Now, I don’t want any of this used to jack up prices by essentially forming a cabal, an anti-competitive cabal. So you’ve got to have safeguards in place, but I don’t want legitimate concerns about liability to stop these companies from doing what they should be doing, and some of that will be with each other.

The other realignment that I’m utterly fascinated by is around privacy and specifically Flock cameras, Axon cameras; people hate them. They hate them so much they’re kicking them out of their communities. We’ve been trying to get the CEO of Flock on the show. I think his company’s in a little bit of a crisis. We’ll see if we can reel him in to talk about what’s actually happening to his business as he faces bipartisan backlash.

I think Americans are redefining what privacy means. Famously, there is no expectation of privacy in public. This has been foundational in our law, and that’s because not everybody had an always-on camera. They certainly didn’t have cameras in their glasses. They didn’t have watches that could record summaries of every meeting they’ve been in all day. And the state wasn’t putting up cameras everywhere that could track all of your movements across every state you’ve ever been in, with a giant database that was enabled by AI so the police could search it easily.

That’s a big swirl, but to me it feels like there’s an assertion of a privacy right in public by the American people. They want to have some kind of privacy in public that does not exist in our law. So society is moving ahead of the law. How would you square that? How would you be able to write a piece of legislation that affords people some sense of privacy in public?

I think the Flock cameras are really disturbing. And yes, there are legitimate law enforcement uses, and they’ve helped catch people who’ve committed terrible crimes, but they’ve also allowed a department to try to follow [people] through their license plate. Someone wrote an op-ed criticizing a police department’s use of these Flock cameras. They’ve also been used to try to track a woman who needed reproductive healthcare. They’ve been used for a lot of illicit purposes, and I think people are waking up to all the problems that they’re creating and all the privacy that’s vanishing. 

Here, and I say this somewhat hesitantly because there’s a lot to be worried about with AI, but this is a cardinal concern and risk with AI, and that is: can privacy exist in the era of powerful AI? We talked a bit earlier about open models being available for defense as well as offense, and to be able to respond in real time. As those models get more powerful, and even the closed models get more powerful, you can train those models to invade people’s privacy. You can train those models to find out what your health records are and what you watch on TV. You can train those models to turn on the microphone on your watch or to turn on the microphone on your Alexa device when it’s turned off to aggregate all that data about you. That’s quite terrifying. 

We may have no privacy left unless we act to ensure that we have privacy. And even then, foreign bad actors will have access to these very sophisticated models. They’ll develop their own. They’ll distill ours. They’ll do all kinds of stuff so that they can have access to it, and they’ll be private sellers, maybe not American private sellers, but other private sellers. We see this in some of the private sellers of these hacking technologies already, some of these companies that are providing eavesdropping capabilities to the highest bidders.

So we’re going to have to figure out how we not just prohibit conduct, but how we beef up the privacy of American people. How do we use technology to protect our privacy? Otherwise, there’ll be little of it left.

The reason I’m asking about what, to me, feels like bipartisan issues to the point that there is a realignment of our politics, is because we’re a month away from an election. It feels, based on the polling, that Democrats are going to win a healthy number of seats, if not outright take control of both houses of Congress.

What should we expect to see if that happens? Is it a bunch of investigations? Is it actually getting the CEO of Flock to show up to a Senate hearing? They didn’t do that a couple of weeks ago. Or is it a flurry of litigation that lands on President Trump’s desk that ultimately goes nowhere?

You’re going to see vigorous oversight in this area to begin with. We’re going to want the leaders of this industry to come in and talk about all of the times that their own agents have gotten away from them, how it happened, what’s been done about it, and what they’re doing about it. We’re going to want these companies to be transparent with Congress about it. We’re going to want to legislate in this area, and that may come into conflict with the president, but there’s a lot of bipartisan work product here. And I think that given where the country is and where it’s moving, it becomes increasingly difficult for this president to veto bipartisan work product coming out of the Congress. 

So that is my hope. It’s certainly a far better prospect than we have today. This is going to be a very central priority for the American people. We want to harness AI for the good it can do. We want to use it to bring down the cost of things. That’s our central priority. And there are ways that AI might be useful in bringing down certain costs. We want to protect the dignity of work also, but we want to make sure that this doesn’t pose a threat to the American people.

You’ve brought up the corrupting role of money in politics several times so far. It occurs to me that the Trump administration, the president, and his family have done what amounts to crypto scams a bunch of times throughout this administration. The second time around, they’ve enriched themselves in ways that I think a lot of ordinary Americans find to be beyond the pale. You served as impeachment manager for the first Trump impeachment. Do you think it’s strategically wise for the Democrats to show up and try to impeach him a third time over this?

I think we’ll need to see what the priority is at the time we take control of both houses, knock on wood, and what the priority needs to be. And I would put the very top priority as fulfilling the promises that we’re making right now, which is to have a central focus: the highest priority on reducing costs for people. So, in the oversight area, what that means, I think, is doing oversight to make sure that the corruption of this administration isn’t raising costs on the American people. And let me give you one example. I could give you about a hundred. I mentioned the Melania movie. I think the American people deserve to know: Is Amazon getting federal contracts now not because it’s the best technology for cloud computing, not because it’s offered the most competitive price, but because it paid off the First Family $40 million?

If they are, we’re paying for that excess. We’re paying for a product that may not be as good. I’m not saying that this has happened, but they’re not going to do any oversight of it. Their decisions on everything from mergers to government contracts to pardons are influenced by the money they’re getting hand over fist. So we, I think, from an oversight point of view, need to focus on oversight that can help attack corruption that’s driving up costs. In terms of impeachment, I think we’ll have to cross that bridge when we get to it.

I’m glad you brought up the entertainment industry, because I do want to end by talking about Paramount and Warner Brothers, which is a merger happening in your home state of California. It feels like a lot of that merger was the Ellison family in particular making promises about the news coverage of CBS News and CNN in order to get favorable regulatory treatment. We can see that CBS has softened its stance. The market has rejected CBS’s softened stance towards this administration. So there’s at least one check on that power, but California Attorney General Rob Bonta didn’t push all the way to the finish line to try to prevent that merger from happening. Do you think that was the right decision?

I’m still frankly looking at the deal that was struck, and I have a lot of concerns about it. The one that you mentioned is a very significant concern, and that is that there was some editorial independence board that’s supposed to oversee CBS and CNN, and yet the Paramount company, or the merged company, the board appointed by the Ellisons, will agree on what that board is and how it’s constituted. So is this real, or is this going to be fiction?

I’m also concerned, frankly, about what it’s going to mean for jobs. I live in Burbank. It’s the entertainment capital of the world, and represents so much of that industry in California. And what we have seen in past mergers is thousands and thousands of layoffs. So that is a particular concern for me. I love this industry. I love movies and television. It’s something I really deeply care about. I’m carrying a bill right now, important to the whole industry, and that is a film and television tax credit to try to bring production back to the United States. There’s never been a more difficult time to work in that industry, but I do have continuing concerns about editorial independence and also the degree to which CBS paid off the president, I think, because they had mergers pending. And ABC paid off the president. Those were personal sums to the president and his pet causes, with the result of completely frivolous litigation by the president that they would have won in court, which I urged them not to settle because of the destructive First Amendment precedent.

The idea that the Trump-appointed FCC chairman attacks ABC because they don’t like the jokes that Jimmy Kimmel tells, or cost Stephen Colbert his job on CBS because they didn’t like the jokes he’s telling — that’s third-world dictator kind of stuff. That’s not what we are used to seeing in the US, and I never want to see it here.

The jobs piece is really interesting to me. One of the longest-running tropes at The Verge is that a workable antitrust policy for the United States would be simply to forbid companies from buying Time Warner because it has never worked and it may never work again, and that we should just pass one law. Just don’t do that, and maybe everything will be fine. 

Paramount made a bunch of noise. They fulminated about leaving California if they weren’t allowed to pursue this merger. They pretended to investigate Nashville and Texas and all the other places that you investigate, and then the merger was approved, or the litigation dropped, and I think David Ellison put out a statement like, “California is our home. We’re going to invest here,” and it seemed very fake in the end. Do you think that threat to leave spooked Gavin Newsom, spooked Rob Bonta?

I think the threat to leave California concerned a lot of people, and I’m sure it concerned them. Was it a real threat? I don’t know. We’ve seen some of these corporate headquarters moves mean that they move basically their office, but they don’t necessarily move the studios or the lots, which would be difficult and expensive for them to do for a whole lot of other reasons. But just the idea, frankly, I found offensive: “If we don’t get approval of the merger we want, we’re going to pick up, and we’re going to leave the state.” It doesn’t say much about a commitment to the workforce if you’re ready to do that at the drop of a hat. So those arguments, really, I found offensive: the claim, “We’re going to leave if we don’t get what we want.”

I will say this: the industry’s been through a hell of a time. The workforce has been through a hell of a time. Between the pandemic, the strikes, the consolidation among the streaming companies, and now the mergers, there’s never been a more difficult time to work in the industry. And I’m not talking about the Tom Cruise-level celebrities. I’m talking about the carpenters and the set designers, the wardrobe people, the dry cleaners, the caterers, the small restaurants and small businesses that thrive when you’re making a film or TV show in LA or somewhere else, that just take a huge hit when they’re not filming. Those are the people that I care about. And we do need stability, so this needs to be resolved. We need stability. We also need to match the incentives other countries are giving if we want this industry to thrive again.

Let me bring this full circle to AI. We talked at the very beginning about the AI companies just ingesting all of this creative output without any payment, and that has enabled the models to be as good as they are. The models are now good at film production. I will admit to having watched several AI microdramas on TikTok, mostly by accident, but they were compelling enough that I stayed. And then, as editor-in-chief, I try to assign our reporters to write about AI microdramas. They did not like this idea, but that’s what happened. That is killing them at the input stage. We’re not paying them for the training material. And now we’re competing in a marketplace with a vastly cheaper product. You don’t need carpenters and small businesses to make AI slop microdramas on TikTok.

How would you connect these dots? Can you make AI compatible with the film industry in California as it exists today?

You know, a lot of these things all come back to the same thing, and that is: what is the commitment to the workforce? Does that matter to these companies? This is a broader problem. I remember having somebody in the industry, a worker in the industry, come to visit me to basically complain about the Affordable Care Act and how it had cost him his job. And I said, “Come by, I want to talk to you about this.” And he came in, and he explained that the studio that he worked for, every year, they would ask for pay increases, and they were told they couldn’t give them more of a pay increase because healthcare costs were rising so fast and it was putting downward pressure on wages. And then the ACA passed, and then he got laid off. And he assumed it was because of the ACA.

I remember telling him, “Honestly, I don’t think it was the ACA. You’re right about healthcare costs going up,” and they were going up very fast. Actually, after the ACA, they were still going up, but they were going up at a slower pace than they were before the ACA. “Here’s what I think happened to you.” And what had happened around the time he was laid off was that there was a hostile takeover bid mounted against one of the studios. It was unsuccessful, but the studio that was the target of that takeover shed all the jobs it would have shed had it been taken over, so it would be a less attractive takeover target in the future. And then other studios shed jobs, even though they were very profitable; they shed jobs because they didn’t want to be taken over. And so there were these cascading sets of layoffs in the industry, none of which were necessary because they were profitable companies, but it did increase the quarterly report, the value of the quarterly report, the next shareholder report.

How do we get out of the bind where the only consideration is the shareholder report and the employees simply don’t matter? Unless the employees start to matter to these companies, then yeah, they will use every device they can to make films without having to use human beings. And this is, I think, a broader challenge in our economy. We see other countries like Germany doing a better job. They have employee representatives on their board of directors, so they take into consideration the impact on their own workforce. You mentioned the prior FTC chair. She was committed to looking at factors in terms of competitiveness and anti-competitiveness, including the impact on the workforce. And I think we should, but this is a continuing challenge.

You’re not up for reelection in this cycle. You’re not up again until 2030, so you have a little bit of distance from the actual campaigning. We’re seeing a bunch of progressive candidates on the left propose muscular reforms to capitalism, right? Abdul El-Sayed has said, “I’m a capitalist, but we need to regulate capitalism so it works for people.” I live in New York. Mayor [Zohran] Mamdani is here. He wants to regulate Amazon delivery trucks.

There’s a lot of that energy on the left. Do you think if the Democrats retake Congress, we’ll see that turned into results on any meaningful time frame, or do you need the presidency too?

I think what we should do is try to get legislation passed on a bipartisan basis because that’s the only way you can get bills through Congress, particularly if you don’t control the White House. But I also think we need to take up big Democratic priorities that may get through the Senate or may not, depending on whether we flip the Senate and what our margin is; they may get vetoed by the president, but show the country that this is the kind of bold reform that we are prepared to pass if you give us the presidency two years later. 

And it is simply not going to be acceptable to anyone to say, “Hey, if you elect us, we’ll take you back to the status quo before Trump came along.” No, the reason Trump came along is people were deeply unsatisfied with the status quo. So we need big, bold ideas. I favor Medicare for All. I’ll let anyone who wants to participate in Medicare participate in Medicare. I don’t think we should force people into Medicare if they want to stay in a private plan, but I think we ought to broaden Medicare and make it accessible to people. It’s a very efficient healthcare system. And that is different than simply saying, “We’re going to restore premium tax credits under the ACA.” Yeah, we should do that too, but that alone isn’t going to cut it.

So I hope we’re going to think big. And I’ll say this too. The reason why we keep going back and forth between a Trump and a Biden is that people are so unsatisfied with how hard they’re working and how little they have to show for it. The economy’s just not working for people. It’s not that they’re out of work. The problem is that they are working and they still can’t make ends meet. And until we address that, our democracy is not going to be on solid ground. It’s going to always be ripe for any demagogue who comes along promising they alone can fix it. We have to make sure that if you’re working hard in this country, you can provide for yourself a good living and for your family.

Last question, and it’s directly related to that. I’ve said several times now that I think we’re in a generational political realignment. It’s been the foundation of many of the questions I’ve asked. Do you agree that we’re in the middle of a foundational political realignment?

We, I think, have been for some time. I’m not sure where the alignment goes from here, but I’ll tell you where I’d like it to go from here. We, in the Democratic Party, we’re unquestionably the party of working people and average families. We’re the party of the underdog, and I think we’ve lost some of that in the public perception. Our policies are still very much driven that way, but increasingly people just think, “Both parties are corrupt. Neither party’s really looking out for me.” I want the Democratic Party to once again be the party that the public looks to and says, “Okay, they’re fighting for me. They’re not fighting for those who’ve already done great. They’re fighting for me.” That’s why I was drawn to the Democratic Party.

And I also think, and this is related to the whole conversation we’ve had, wealth is now concentrated so much in so few hands that it is totally warping our political system. Wealth is perpetuating itself and warping our system even further. Some people took it as a celebration: we have our first trillionaire. Well, that first trillionaire, Elon Musk, became a trillionaire because of government largesse, because of government contracts that saved his bacon when his companies were in trouble. And there is no ethics, no sentiment about giving back there. No, he was only happy to DOGE the people who made his success possible.

But what made that possible too, that is the aggregation of all that wealth, is a tax system that’s not the least bit progressive anymore and is becoming less and less progressive. And one of the missions I hope for our new Congress, and I hope for the Democratic Party, is: how do we have a truly progressive tax code again? How do we make sure that, yes, the AI companies pay their fair share and they pay for some of the disruption that technology is going to cause? But I think even more important, how do we make sure the tax code is truly progressive and people who are doing well are going to pay well into the system? That, I think, is fundamental.

Senator Schiff, this has been a great conversation. Thank you so much for being on Decoder.

Great to be with you. Thanks for having me on.

_{Questions or comments? Hit us up at decoder@theverge.com. We really do read every email!}