Lauren Feiner | The Verge

Democrats think cracking down on AI could give them an edge

2026-10-19T08:37:05-04:00

Democrats could win back control of Congress this November, and if they do, they’ll need to decide how to approach one of the country’s newest and thorniest political issues: AI.

Polls predict Democrats will flip at least one chamber of Congress in the midterm elections, which would put committee leadership back in their hands. That could usher in a wave of newfound scrutiny for AI companies. Top Democrats on key committees have already signaled their willingness to press the industry on tax credits tied to data centers and recent hacking incidents, and would be able to set committee agendas if they take over, allowing them to call CEOs to testify in Congress and demand documents that explain companies’ inner workings.

AI issues are already taking a leading role in campaign ads, particularly as data center development has driven residents across the political spectrum to packed local meetings. While President Donald Trump has opposed most regulation of the technology in favor of an accelerationist approach, public opinion polling has shown that Americans have widespread concerns about how AI will impact jobs, education, and other aspects of life. The direction and intensity of that oversight, though, is up for grabs.

The issue of AI is “a hundred percent ready for the taking,” says Jesse Lehrich, a Democratic strategist who previously ran a tech accountability advocacy group and worked for Hillary Clinton’s presidential campaign. “The question remains open as to whether Democrats are willing and able to seize that opportunity.” AI issues currently include a lot of different things, from monitoring and restricting the development of frontier models to dealing with the effects of job displacement or data centers’ energy demands. Across both parties, lawmakers are still figuring out where to put their stakes in the ground, and early congressional proposals have yet to gain broad consensus.

“The question remains open as to whether Democrats are willing and able to seize that opportunity”

AI caught many lawmakers off guard, says Rob Flaherty, former deputy campaign manager for Kamala Harris’ presidential campaign and a Biden White House alum who cohosts a podcast with Lehrich. “For a long time, Democrats were afraid to talk about it, either for a mix of not understanding it, or being afraid of the money that was going to be poured in by the AI industry,” Flaherty says. But as AI becomes more salient for voters, “there will be immense profit to whoever can speak to the world that we’re moving into at a volume that voters can hear and can feel.”

Flaherty recently launched a super PAC called Tomorrow Together with Karthik Ganapathy, a former state communications director for Bernie Sanders’ presidential campaign. The group launched its first $4 million ad blitz last week in battleground House races, attacking Republican candidates as out of touch on AI and referencing issues like kids’ online safety and data centers. The group is looking beyond the midterms toward the 2028 presidential election, aiming to hone Democrats’ messaging to double down on what resonates most with voters.

While data centers have been the most visible AI issue in many communities, Flaherty says his group’s polling found sweeping bans on state AI regulation — repeatedly pushed by Trump and congressional Republicans — are even more broadly unpopular. Democratic donor and venture capitalist Ron Conway now regrets funding Marc Andreessen-backed pro-AI super PAC Leading the Future, Jay Carney, a former Obama press secretary who now runs communications for Conway’s VC fund, told Politico.

As of June, LTF spent nearly $45 million over almost a year, with $31 million left in the pot according to Federal Election Commission filings. “At a certain point, voters speak louder than money,” Flaherty says. Even governors who lean toward the center of the Democratic Party, like California’s Gavin Newsom or Pennsylvania’s Josh Shapiro, have taken bolder stances against AI in recent weeks. Newsom issued an executive order looking to create an AI “kill switch” after previously vetoing a bill that would have instituted some similar measures, and Shapiro reversed course against data centers.

Democrats are well positioned to capitalize on the negative sentiment toward AI

Democrats are well positioned to capitalize on the negative sentiment toward AI because Trump’s largely anti-regulation stance has boxed in his party. Though some Republicans, including Florida Gov. Ron DeSantis and Sen. Josh Hawley (R-MO), have carved out a more skeptical approach to AI, Lehrich says, “I cannot think of a bigger political gift than Donald Trump just wholly bear-hugging the industry and declaring that we must let data reign and any community that rejects data centers is backwards and poor.”

A more sweeping AI regulatory agenda might include embedding independent researchers into frontier labs, requiring more safety checks up front, and imposing more liability on tech companies for anything from how their tools affect kids to whether they’re used to discriminate against workers or housing applicants. Lawmakers could also consider ways to meet demands from constituents that AI companies help pay for job displacement and other impacts they’ll have on people’s lives.

“I cannot think of a bigger political gift than Donald Trump just wholly bear-hugging the industry”

AI is still not the top issue driving voters to the polls, but Flaherty says that’s not necessarily the right way to think about it. “AI is not an issue as much as it is a story,” he says. “It’s a story of politicians getting bought off by industry. It’s a story of a bunch of guys making you foot the bill for them getting rich while they’re coming to make your life worse. It is a story that fits the national vibe.”

Where exactly different policy lines should be drawn on AI is still somewhat unsettled. But Flaherty warns that Democrats risk giving up ground if they fail to take big enough swings. Voters are “giving politicians a lot of leeway to take big swings and to really reimagine stuff,” he says. If politicians only stick to basics like plans to retrain workers at community colleges, “I think we’re going to miss the bigness and the ingenuity that the moment calls for.”

Now that Character.AI is going bankrupt, what happens to all of its users’ data?

2026-10-15T10:12:33-04:00

New York Attorney General Letitia James revealed a years-long multi-state investigation against Character.AI as they now seek to require protections on consumer data sold through its bankruptcy proceeding.

James is among 40 bipartisan AGs who filed a notice of interest in Character.AI’s bankruptcy case “to put all parties on notice of the regulatory issues that will need to be addressed in any sale or transfer of the Debtor’s business or assets.” Character.AI announced Tuesday it was filing for Chapter 11 and selling some assets to Disney. But the AGs warn there could be serious risk involved in the transfer of the technology due to the “highly sensitive nature of the conversations many users have with Character.AI chatbot characters, the large number of minors who use the product, the serious risks posed by specific features of the product.”

The filing also reveals that several of the states had been mid-investigation when Character.AI filed for bankruptcy. The states say they opened a probe into consumer protection and privacy concerns in 2024, following a lawsuit from the mother of a 14-year-old who died by suicide allegedly “seconds” after talking with the chatbot.

“The investigations uncovered exceptionally serious dangers”

“The investigations uncovered exceptionally serious dangers and potentially unlawful conduct from Character.AI, particularly its treatment of children, including chatbots posing as licensed professionals, exposure to sexual and violent content, weak age checks, misleading safety claims, and potential violations of state privacy laws,” according to a press release from James’ office. The states were working to address these issues with Character.AI, but didn’t reach an agreement before the company filed for bankruptcy. In a statement, Character.AI CEO Colin Adams said the AGs’ filing “is not an accurate portrayal of the safety of our platform. Character.ai takes the safety of its users and its platform extremely seriously, especially as it relates to minors. Character.ai has and continues to monitor its platform to prevent and remove content that violates its standards and policies.” Disney did not immediately respond to a request for comment.

The states are urging the judge to require the purchaser of Character.AI’s assets to implement procedures that will safeguard the sensitive information the company maintained. Those include effective age-assurance systems, restrictions on how minors can engage with the chatbots, implementing “Crisis safeguards,” and letting users delete their data. “This company collected some of the most private thoughts children have ever shared, and now those conversations could be sold off to the highest bidder,” James said in a statement. “I am fighting to make sure any buyer puts children’s safety first and puts real protections in place.”

Update, October 15th: Added statement from Character.AI.

Frances Haugen hopes The Social Reckoning will inspire more whistleblowers

2026-10-09T12:50:40-04:00

NEW YORK, NEW YORK - OCTOBER 04: Frances Haugen attends the Social Reckoning world premiere at Regal Union Square on October 04, 2026 in New York City. (Photo by Sean Zanni/Getty Images for Sony Pictures Entertainment) | Getty Images for Sony Pictures E

Facebook whistleblower Frances Haugen was mostly unfazed watching Oscar-winning actress Mikey Madison play a character named after her in The Social Reckoning, except for one thing: the wrist warmers.

In Aaron Sorkin’s new thriller, the fictional Haugen — who, like her nonfictional counterpart, leaks thousands of documents to Wall Street Journal reporter Jeff Horwitz — habitually dons the accessories. It’s a character quirk that reflects real life, where they’re a way to deal with the effects of a vasoconstrictive disorder. “I’ve figured out lots of ways to minimize how much of an obstruction this is,” she tells The Verge in an interview. “But having to watch myself wear wrist warmers for two hours, I came out of it being like, ‘Oh my god, this is really disruptive. Like, how have I successfully not thought about this for like seven years?’” With The Social Reckoning appearing in theaters Friday, Haugen hopes Meta’s CEO might watch it and have a similar moment of recognition with his own counterpart: a callous, robotic character played by Jeremy Strong. “I hope Mark Zuckerberg watches the movie and has as much of a knee-jerk emotional response as I had on that regarding, ‘Do I really want to live this life anymore?’”

Strong’s version of Zuckerberg has become sick of taking criticism for content moderation and fact-checking decisions as he prepares for congressional testimony, and waves off concern for design choices that his staff believe will stoke anger on the platform. “It’s very possible that Mark has learned to not see how distorted his world has become,” Haugen says. “My hope is that holding that mirror up, if he does ever watch the movie, that it gives him the gift of perspective, so he at least has a chance to choose.”

“I hope Mark Zuckerberg watches the movie and has as much of a knee-jerk emotional response as I had”

Five years since she went public as the figure behind the Journal’s Facebook Files, Haugen hopes that the film can remind viewers about the breadth of information they uncovered and “humanize the whistleblowing experience.” While she says her mother may take issue with Madison’s young, brunette character being shorter and less blonde, she’s happy the fictional Haugen stands apart from her. “We want as many people as possible to see themselves in Mikey’s character,” she says. If someone at a frontier AI lab becomes “so unlucky that they find themselves in a place where they need to blow the whistle,” she hopes they’ll connect with the fictional Haugen and “be like, ‘Oh, I actually have a duty to my common human beings and fellow citizens to do that.’”

One of Haugen’s most tangible impacts was providing evidence in lawsuits that claim Meta harmed young users, culminating in a $17 billion settlement earlier this year. But as the movie reminds viewers, child safety wasn’t what initially motivated her. In fact, documents about Instagram’s impact on teens were among the last she collected before resigning, because she worried clicking around too far outside her own subject area of civic integrity would raise alarms. “The collective memory of ‘What exactly was I complaining about?’ has shifted,” Haugen says. “Only about 10 percent of what I brought out was about kids.”

Revelations about how the algorithm pushed users toward extremist groups, special content moderation privileges for elite users, and other issues “kind of disappeared into time,” she says. Meta declined to comment on the movie but has previously said the documents Haugen disclosed were cherry-picked and flatten decisions the company has had to make amid difficult tradeoffs.

“We want as many people as possible to see themselves in Mikey’s character”

Still, Haugen has since engaged frequently on policy around social media and kids, much of that in Europe and Australia, where movement seemed more likely than in the US. While she says she’s “not a prohibitionist” when it comes to social media, she believes Australia’s youth social media ban pushed Meta to implement what she calls “‘yeah, duh’ features” to combat the threat. “I don’t think the first line of defense is you ban kids off social media,” she says. “But the problem is that platforms have shown repeatedly that they’re not particularly interested in making these spaces safe for kids, and Facebook particularly has shown they only improve safety for kids when they’re scared.”

Meta has aggressively moved to silence another recent whistleblower, former public policy director Sarah Wynn-Williams, taking legal action to stop her from promoting her memoir about her time at the company, Careless People. But Haugen says she sees Meta’s action against Wynn-Williams as an “outlier,” in part because her retelling relied much more on personal narrative than Haugen’s reams of documents. Before Haugen’s character obtains the documents, “there’s a scene in the movie where the Jeff character tells the Frances character, ‘You didn’t give me evidence, you gave me a story,’” she says. “And because a lot of what [Wynn-Williams] is saying is dependent on her personal credibility, Meta very, very desperately needs no one else who had access to that same information to ever come forward.”

As hype over artificial intelligence overtakes Meta and the rest of Silicon Valley, Haugen says conversations around industry regulation and public dissent are already much further along in AI than they were at a similar point with social media. Though she says some proposals seem “simplistic or misdirected,” like a pause on development, she’s encouraged that “we see people engaging, and we did not see people even engaging four years into social media.”

Reflecting on her own decision to go public as a whistleblower, Haugen tells The Verge that she took a pretty dispassionate approach. She knew Meta could cause financial hurt, but says she didn’t worry too much about that. “I’m a Quaker and I don’t take a lot to be very content and happy,” Haugen says. “They can only take as much from you as you let them.”

What The Social Reckoning gets right — and wrong

2026-10-07T14:00:20-04:00

Image: Dimitrios Kambouris/Getty Images | Getty Images

Aaron Sorkin’s The Social Reckoning has its fair share of cinematic flair, but many details in the story hew closely to real life. The movie tells the story of how Frances Haugen, a Facebook civic integrity product manager, came to leak a large trove of internal documents to Wall Street Journal investigative reporter Jeff Horwitz, in what would later become a 2021 series called “The Facebook Files.” In both real life and in the film, Haugen’s leaks lead her to testify as a congressional whistleblower. The final frames of the movie describe the impact she had even after that closing scene, speaking before governments around the world and bolstering lawsuits against Meta nationwide with the internal communications she disclosed.

The film’s Haugen, played by Mikey Madison, says that she really believes in Facebook, echoing a message the real Haugen left for her colleagues just before she signed off for her last day in May 2021. “I don’t hate Facebook,” she wrote on the company’s internal message board, according to The Journal. “I love Facebook. I want to save it.”

Sorkin, both directing and screenwriting the semi-follow up to 2010’s The Social Network, which he only wrote, took some liberties with the dialogue and the chronology of the movie. The most obvious is that he condensed the timeline and shifted major events forward by several months. Some key moments in the film — Zuckerberg (Jeremy Strong) preparing for a congressional hearing about misinformation and extremism and Horwitz (Jeremy Allen White) getting ready to publish the first Facebook Files report — occur on the same day, which happens to be January 6th, 2021, amid the insurrection at the US Capitol. In reality, Zuckerberg testified in the House alongside other tech CEOs in March 2021, and the first installment of the Facebook Files wasn’t published until that September.

“I love Facebook. I want to save it”

But the movie pulls directly from Horwitz’ investigation and the documents that Haugen collected, which showed how the company applied its content moderation policies in inconsistent ways, drove users toward hate groups, and had meager responses to apparent human trafficking on the platform. Meta declined to provide a comment on the movie, but in past statements, the company has referred to the documents collected by Haugen and their recounting in the Facebook Files as cherry-picked to serve a specific narrative. Zuckerberg called the reporting Facebook Files “a coordinated effort to selectively use leaked documents to paint a false picture of our company” during a 2021 earnings call after the initial reports were released, and said that when faced with “impossible tradeoffs … these questions are not primarily about our business, but about balancing different difficult social values.” Here’s what you should know about the real life events underpinning the film.

Losing a friend to the misinformation rabbit hole inspired Haugen

The Social Reckoning opens with a young blonde woman standing next to her overheating car on the side of a highway. When cops pull up to ask if she’s okay and smell alcohol on her breath, the exchange quickly grows tense. The woman appears unnerved by the officer, a person of color, and whips out her phone to record him; soon, she’s threatening violence and is arrested. Then, we see Madison’s Haugen in her apartment watching the video the woman posted to her Facebook page, “Voice of Reason,” with the caption “Watch me own DEI cops.” Haugen reveals later that the woman was a friend she met in spin class, radicalized by conspiracy theories on Facebook after the progressive candidate she campaigned for lost.

The scene and details about the friend appear to be fictionalized, but are based on a true part of Haugen’s story. Haugen told The Journal that she took the role on Facebook’s civic integrity team after seeing a family friend radicalized online. “It’s one thing to study misinformation, it’s another to lose someone to it,” she said.

Facebook’s algorithm recommended extremist groups

In the movie, one of the first major revelations Haugen shares with White’s Horwitz is that 64 percent of people who joined extremist groups on Facebook were sent there by its algorithm. That statistic — in the movie and real life — became the basis of a May 2020 article in The Journal. In the real article, Horwitz and co-author Deepa Seetharaman wrote that a 2016 internal Facebook presentation found “64% of all extremist group joins are due to our recommendation tools.” It added, “Our recommendation systems grow the problem.” At the time of the 2020 article’s publication, a Facebook spokesperson told The Journal it had since strengthened protections and “are not the same company today.”

A shift in Facebook’s algorithm promoted rage

As they hunt through pages of documents in the movie, Haugen and Horwitz keep trying to figure out why Facebook tuned its recommendation algorithm to reward “Meaningful Social Interactions” (MSI), rather than likes, in 2018. Haugen senses favoring MSI could be driving users like her friend toward radicalization, but she and Horwitz initially aren’t sure how — especially since Facebook publicly said the shift was expected to undercut engagement. They come to understand that the new ranking system weighted emoji reactions and reshares much more highly than likes; Zuckerberg and other executives knew this change would promote rage-sharing.

That’s a fairly accurate translation of what we know about MSI. Zuckerberg promoted the shift as a way to correct for a trend of “public content” from brands and media “crowding out the personal moments that lead us to connect more with each other.” Though he anticipated the change would be good for users, he said it would actually hurt Facebook’s engagement. But, he said, “if we do the right thing, I believe that will be good for our community and our business over the long term too.”

“Misinformation, toxicity, and violent content are inordinately prevalent among reshares”

In an installment of The Journal’s “Facebook Files” series in 2021, however, Horwitz and co-author Keach Hagey found that Facebook employees had warned that prioritizing MSI was promoting angrier engagement. Documents obtained by The Journal showed that leading up to the change, Facebook had been concerned about declining engagement from users through actions like resharing or commenting. To combat this, the documents showed, Facebook tried to encourage more original posts and resharing. While Facebook found some positive user benefits of the change, researchers found, “Misinformation, toxicity, and violent content are inordinately prevalent among reshares.” Zuckerberg, in the movie and apparently real life, resisted some mitigation proposals that would damage engagement.

Meta has downplayed the significance of MSI. In a statement to The Journal at the time, Meta spokesperson Andy Stone said, “Is a ranking change the source of the world’s divisions? No.”

Instagram made teen girls feel worse about their bodies

In the movie, Haugen’s colleague shares a thumb drive with documents including a study about how Instagram harms teen girls’ body image. She tells Haugen to get the word out before the company starts marketing Instagram Kids. It’s not clear if this colleague actually existed, but the information she shares appears to be based on an internal study that found 32 percent of teen girls surveyed who said they felt bad about their bodies felt worse after using Instagram. Instagram had indeed planned to launch a service specifically for kids under 13, but announced shortly amid the uproar from the “Facebook Files” release that it was pausing the project to work through concerns.

An early Facebook investor encouraged Zuckerberg to change

As Strong’s Zuckerberg preps for the 2021 congressional hearing in the film, sitting on a stage behind a desk and rehearsing answers with his team, a venture capitalist named Charlie emerges from the shadows. Charlie, played by comedian Bill Burr, gives Zuckerberg feedback and urges him to moderate political ads. Movie-Zuckerberg says he’s upset with Charlie after he wrote a scathing op-ed, and reminds him he 100x’d his investment, which would have been worth over $1 billion if he’d held his position.

Charlie is inspired by real-life Facebook investor-turned-critic Roger McNamee, an executive producer of the film. McNamee published an op-ed in The Guardian in 2017 charging that Facebook and Google “threaten public health – and democracy,” and in 2019 published the book Zucked: Waking Up to the Facebook Catastrophe. Sorkin told Vanity Fair that although McNamee was game to be depicted in the movie, “it was made clear” that “there might be a problem with Facebook” in depicting the conversation between the VC character and Zuckerberg in the movie. So the character became a fictionalized version inspired by McNamee, instead.

Facebook’s civic integrity team was “absorbed”

In the film, Haugen seems dejected when she realizes that the civic integrity team she works on — which worked on worldwide election-related risks including misinformation — will likely be dissolved, and she’ll be reassigned. A Facebook executive assures the team that even if that were to happen, the work would be “absorbed” into other teams, and it would only be beneficial to their work.

“I don’t trust that they’re willing to actually invest what needs to be invested to keep Facebook from being dangerous”

This reflects how the company communicated the change in structure following the 2020 US election. In her “60 Minutes” interview in 2021, Haugen said, “when they got rid of Civic Integrity, it was the moment where I was like, ‘I don’t trust that they’re willing to actually invest what needs to be invested to keep Facebook from being dangerous.’” (That’s actually when Haugen first responded to Horwitz’ initial outreach on LinkedIn.) Meta executive Guy Rosen responded at the time that the team was not disbanded, but rather “integrated” into “a larger Central Integrity team so that the incredible work pioneered for elections could be applied even further, for example across health related issues. Their work continues to this day.”

Zuckerberg’s office was bulletproofed

Haugen’s film character mentions that Zuckerberg has bullet-proof features in his office, which appears to be based in reality. A 2019 report from Business Insider found that a “glass-walled conference room” he would sometimes work from had “bullet-resistant windows and a panic button.”

It took time for Haugen and Horwitz to build a strong source-reporter relationship

The movie depicts a push-and-pull relationship between Haugen and Horwitz, as the journalist tries to get Haugen to share more information without crossing a delicate ethical line of soliciting her to steal confidential documents. As Haugen decides whether or not to go the extra mile, she also becomes frustrated with Horwitz for not making more progress on his own.

The two meet regularly on a hiking path, where Horwitz eventually gives Haugen a burner phone to communicate with him and on which she later takes pictures of internal documents. At one point, Horwitz even travels to Puerto Rico to collect more information from Haugen, after she relocated there during the pandemic.

These details all seem to mostly follow the true course of their source building relationship. Haugen told The New York Times in 2021 that she had “auditioned” Horwitz for some time, and decided to give him the scoop because she found him to be “less sensationalistic than other choices I could have made.” Horwitz acknowledged in an on-stage interview with Haugen in 2022 that “things got tense between us” at times, in part because Haugen had concerns about The Journal’s institutional approach, and in part over some messiness about when certain information was released by other publications or agencies Haugen worked with.

Haugen really did relocate to Puerto Rico during the pandemic, and Horwitz also visited her to collect information, though the timeline of those travels is different than in the movie. Haugen invited him to Puerto Rico once she realized Facebook’s policies prevented her from working from a US territory, prompting a deadline to quit or move. She chose to quit, meaning the pair had one last period of time they could access internal documents at Facebook. Speaking to Horwitz on stage in 2022, Haugen said she was “so grateful that you put in a huge amount of effort to just be there and make sure that we asked as many questions as we could so history got to have some kind of record for it.”

NYC is now the first city in America that bans sketchy subscriptions

2026-10-01T11:12:29-04:00

Image: Angela Weiss / AFP via Getty Images | AFP via Getty Images

New York City residents struggling to get out of recurring subscription fees can now submit complaints to the city government. As of Thursday, the city’s click-to-cancel rule has taken effect, which requires businesses to make it as easy to cancel a subscription as it is to sign up. That means a gym that lets you join online, for example, can’t make you call or go in person to end your subscription.

It’s the first municipality to see the rule take effect, after an appeals court struck down a federal version for procedural issues. The city expects it to save New Yorkers anywhere from $21.5 million to $162.5 million annually. And to NYC Department of Consumer and Worker Protection (DCWP) Commissioner Samuel Levine, it’s an example of how state and local governments can hold companies accountable when federal efforts fail. “No one should need 45 minutes of hold music to stop paying for something they never wanted,” NYC Mayor Zohran Mamdani said in a statement. “If a company can take your money with one click, you should be able to get your money back with one click. And now, if they won’t let you cancel, the City is coming for them.”

Levine was the Federal Trade Commission consumer protection chief under Lina Khan, and he tried to usher in the click-to-cancel rule at a national level. He says the experience of implementing the rule in NYC has been “night and day.”

“We proposed this rule this year. We finalized it this year. It’s taking effect this year. This is the speed of government as it should be,” Levine tells The Verge in an exclusive interview. While Levine says he’s proud of his work on the FTC rule and sees it as a “model” for other state and local governments, “at the end of the day, it took more than four years. And then after all of that, it never actually gave Americans the relief they needed.” While the FTC under current Chair Andrew Ferguson has taken an initial step toward potentially reestablishing a similar rule, Levine is betting that more municipalities will do so first. While that leaves fewer Americans protected, it still creates a deterrent for businesses to engage in deceptive subscription practices.

“This is the speed of government as it should be”

In New York, consumers can use a new portal to file complaints if they believe a business gave them unclear subscription terms, delayed cancellation, or failed to inform them the subscription would auto-renew. After filing a complaint, DCWP will review it and could reach out for more information and try to get consumers’ money back and help them cancel relevant subscriptions. If the agency sees a pattern of noncompliance at a certain business, it could prompt further investigation and legal action. Under the rule, businesses could face fines starting at $525 per violation.

The mayor’s Public Interest Technology (PIT) Crew built the portal in 10 weeks; Levine assures consumers that the experience will be far different from the White House’s newly launched America.gov, which he says appears to be “just an AI chatbot layered on top of the federal government’s existing infrastructure.” When consumers submit a complaint to the click-to-cancel portal, Levine says, “they’re not going to be diverted into a chatbot doom loop. We’re going to have actual people whose full-time job is to help them.”

He also says it won’t be like going to a DMV office to renew your driver’s license. “If we want to restore people’s faith in government and the ability of government to deliver, we need to show that government can create excellent products. We need to show that every experience is not going to the DMV. And I mean, no offense to our colleagues at the DMV,” Levine says. “People have really bad associations with interacting with the government, and sometimes that deters them from filing complaints. You can have the best laws in the world, but at the end of the day, if your eyes and ears, the people you’re working to protect, don’t want to actually deal with the government and tell us what happened, those laws are not going to be effective.”

Update October 1st: Added statement from NYC Mayor Zohran Mamdani.

Trump orders US government to call AI ‘Super Intelligence’

2026-09-29T19:34:43-04:00

Image: Carolyn Van Houten / The Washington Post via Getty Images

The US executive branch is no longer acknowledging the existence of “artificial intelligence.” Going forward, official policy websites, policy documents, and press releases will refer only to “Super Intelligence,” thanks to a new executive order signed by President Donald Trump.

“The word super is the best word of all, and it’s the simplest,” Trump said at an event announcing the launch of America.gov earlier on Tuesday, and said Chinese President Xi Jinping, who visited the White House last week, “loves it” too. “We don’t want to hear artificial. Because it’s not artificial. It’s very powerful, it’s very brilliant. It’s going to mostly be put to good use and we’ll stop the bad use.” Artificial, he added, “is like the news. Fake news.” He first mentioned plans for the rebrand in a speech to the United Nations General Assembly last week, aiming to quell fears over AI advancing too quickly and posing a safety risk.

“It’s going to mostly be put to good use and we’ll stop the bad use”

Executive agencies won’t have to change past regulations or documents, but it could still constitute a real (and very strange) shift. Superintelligence is one of several industry terms for particularly powerful versions of artificial intelligence, but the Trump administration is seeking to use the term as a replacement for a broader statutory definition of AI. US officials will be expected not to even acknowledge the use of the terms “artificial intelligence” or “AI” in relevant settings.

Trump signed the executive order after a luncheon he hosted at the White House with tech CEOs and government officials to discuss what he’s now officially calling Super Intelligence. In brief public remarks after the meeting, flanked by Nvidia CEO Jensen Huang and Tesla and SpaceX CEO Elon Musk, and surrounded by other attendees, Trump called the lunch “very productive” and “extremely friendly.”

“There’s a belief that there should be tremendous self-regulation,” Trump said. “And we automatically have regulation with the Department of Justice, the FBI, all of that.” On the local level, where tech companies have faced the most hurdles in building out massive data centers needed to run their services, Trump says the firms will “work to make the community happy” by providing resources and economic support they may need. Ultimately, he says, “data centers are going to be very popular.”

Will Chinese AI companies slow down? A top House Democrat wants answers

2026-09-29T08:00:13-04:00

Image: Tom Williams/CQ-Roll Call, Inc via Getty Images

As President Donald Trump prepares to meet tech and AI CEOs in Washington, Rep. Ro Khanna (D-CA) is calling for a treaty between the US and China to keep AI from wreaking havoc on the world. But wrangling leaders in both countries to take action could be a long shot.

In letters shared exclusively with The Verge, Khanna — the top Democrat on the House Select Committee on China and a possible presidential contender — asked a US intelligence agency and Chinese AI companies whether they would be prepared for an incident like OpenAI agents’ hack of Hugging Face this summer. The companies included DeepSeek, Alibaba, and Moonshot AI, all of which are pursuing high-level artificial intelligence.

American and Chinese interests “could not be more different on many issues,” Khanna writes in the letter to Director of National Intelligence Jay Clayton, “but on this discrete concern, it benefits neither country for humanity to lose control of our destiny.”

“[I]t benefits neither country for humanity to lose control of our destiny”

Khanna requests the Office of the Director of National Intelligence (ODNI) assess how effectively the US might respond to an AI lab losing control of its bots — a scenario that’s become increasingly common in recent months. Alongside it, he asked the agency to evaluate the Chinese government’s approach to assessing catastrophic AI risks. The information, he says, will help Congress and eventually the White House craft an agreement with their Chinese counterparts.

The letters to DeepSeek, Alibaba, and Moonshot AI ask for documentation about their efforts to achieve “superintelligence” and recursive self-improvement (RSI) — a milestone where AI systems can improve on themselves. He asks about any efforts to implement safeguards and “kill switches” for the technology, as well as whether the companies would commit to inspections by a non-governmental agency if a treaty were struck. The companies and ODNI did not immediately respond to requests for comment from The Verge.

Khanna’s move follows a meeting between Chinese President Xi Jinping and US President Donald Trump at the White House last week. Both men, he says, fail to recognize the risk and urgency around AI. “I think it was constructive that Trump and Xi Jinping at least agreed to talk about AI and have ongoing dialogue about it, but it is not nearly adequate,” Khanna told The Verge in an interview on Monday.

He is also skeptical of a planned Tuesday meeting between the White House, House Speaker Mike Johnson (R-LA), and several American tech executives. “I think it’s an effort to reassure the American people that these tech companies are going to be responsible, and it’s a branding exercise,” Khanna says. “But why should we trust companies that have a motive to maximize profits to be making the rules for what is safe? It’s not that they’re bad people, but would you trust the pharmaceutical executives to determine the rules for what drugs are safe?”

“Why should we trust companies that have a motive to maximize profits to be making the rules for what is safe?”

Khanna ultimately seeks a US-China treaty banning RSI and monitoring frontier AI labs, but he recognizes that’s unlikely to happen anytime soon. Sen. Bernie Sanders (I-VT) has similarly called for an AI treaty between the countries, though Republicans like House China Committee Chair John Moolenaar (R-MI) have been skeptical the Chinese government would keep its end of the bargain. In the meantime, Khanna says he’d like a working group with experts from both countries to suggest safety recommendations and methods for ensuring compliance.

But Trump and his allies have been resistant to most regulations or agreements. The morning of his meeting with Xi last week, Trump posted on Truth Social, “Super Intelligence (SI) will be a big topic of discussion, but I want to leave it exactly where it is. That is China’s position also.” Johnson said on Fox Business that he expected the conversation to revolve around striking “the right balance” of oversight without stifling innovation, warning against losing the AI race against China. Leading US AI companies, however, have called for the US to coordinate with China to slow down the pace of development

In the letter to the Chinese AI companies, Khanna acknowledges they’re unlikely to respond in full. He said they didn’t respond to requests in an earlier letter asking for a commitment to slow down AI development alongside OpenAI and Anthropic. Khanna tells The Verge his goal is to “shame them” and “spark a debate” among Chinese AI companies about safety. While he’s critical of US AI firms too, he says, “at least they’re engaging in safety conversations.”

Khanna remains confident the US will drive AI safety standards, in part because he believes many Chinese models are distilled from US models. And while world leaders may not recognize the threat, they’re “out of touch with the scientists and technologists in their own countries,” he says. “We need to get the technologists and the engineers who understand this really driving the debate, not CEOs who have a vested interest in making profits, and not politicians who lack a technical expertise and who are more concerned about power.”

View Link

The vibes are bad for Flock in Washington

2026-09-24T17:53:31-04:00

Flock is in the hot seat in Washington, even if its CEO declined to actually face senators at a hearing about its “AI Surveillance Network” on Wednesday.

“There’s many players in this industry, but there’s one name that really stands out above the rest, and that is Flock,” Senate Judiciary crime and counterterrorism subcommittee chair Josh Hawley (R-MO) said. “If you haven’t heard of Flock, all I have to say is, Flock’s probably heard of you.”

Though the CEOs of Flock and peer companies Axon, Motorola Solutions, and Verkada declined to testify, according to Hawley, seats in the hearing room were in high demand. A standby line stretched down the hallway even after they were filled. Instead of the CEOs, the hearing featured testimony from cybersecurity and legal experts, a sheriff whose department ended its Flock contract after allegedly receiving slippery answers about its AI use, and a Florida woman who said she spent 13 days in jail after her car was incorrectly identified by Flock cameras as being connected to a crime scene. Flock spokesperson Paris Lewbel called her experience “unacceptable, and she has our deepest sympathy for what she went through.” Lewbel said in a statement the incident appeared to be “the result of human errors in judgment unrelated to Flock,” and that her lawsuit against the police department “clearly states that our cameras established she was not, in fact, at the crime scene.”

Verkada spokesperson Virginia Ramsey said the company provided information to the subcommittee to assist in preparation even though its CEO could not attend. “As we’ve explained to the Subcommittee, none of our customers can search a national Verkada network – because none exists. Our customers can only access the data in their own systems,” Ramsey said. Representatives for the other companies invited did not immediately respond to The Verge’s request for comment.

“If you haven’t heard of Flock, all I have to say is, Flock’s probably heard of you”

The federal hearing shows just how much the conversation around Flock has escalated, from concerns within individual police departments about reported abuse by cops, to raucous city council meetings and testy state capitol hearings — and now, a Congressional hearing. Less than two months before Election Day for the midterms, lawmakers across the aisle appeared largely aligned and even chummy over their unease about automated license plate readers (ALPRs) that hoover up location data for Americans who haven’t been accused of crimes, and can store that information, in many cases, for an entire month.

Lawmakers expressed concern about how securely Flock is able to store its vast trove of images, which cybersecurity engineer Benn Jordan said he was able to access with minimal technical maneuvering, in an incident Flock has said it’s since addressed. Jordan said he’s more recently been granted greater access to Flock to evaluate it, and said its security has improved from a year ago, but broader industry-wide issues persist. The lawmakers also took issue with how easy it appears to be for officers to access Flock data, and said it’s distinct from traditional license plate readers, whose systems don’t construct the same dossier of connected information on individual vehicles.

Democrats like Sens. Alex Padilla (D-CA) and Dick Durbin (D-IL) asked how federal immigration authorities could be legally barred from accessing Flock data, and about potential loopholes, which one witness said could be as simple as getting a sympathetic local officer to perform a search on the agency’s behalf. Sen. Cory Booker (D-NJ) asked whether officers in states that restrict abortion access could track people seeking reproductive care across state lines. On the Republican side, Hawley pointed out that similar concerns could apply for gun owners. “For those of us who are supporters of the Second Amendment, we’ve long feared national databases of any kind. You could easily put these cameras up in front of a gun show or near a gun store or a range. You can put them up in front of pregnancy care centers, and you can do all the same things that Senator Booker was talking about,” Hawley said.

“We’re all just getting up to speed on this”

But it’s still early days for legislating around Flock and other ALPRs. After the hearing, Hawley told reporters that he hadn’t heard of the company until a month or two ago, and “we’re all just getting up to speed on this.” He said he’d still like to hear from the CEOs directly, but subpoenaing them to appear would require a full committee vote, and he hadn’t yet discussed that with Judiciary Committee Chair Chuck Grassley (R-IA). Then there’s the issue of whether President Donald Trump, who recently said “I like them” of the cameras, would support legislation that restrains them. Hawley said he hasn’t asked that yet.

“In Missouri, we’ve been using license plate readers forever… but that’s really different than the technology we’re talking about today,” Hawley told reporters. “I think that’s something that I’ve only recently come to appreciate, and I bet most folks are just getting up to speed on. Citizens know though. They know when they’re being tracked. They don’t want to be tracked and treated like criminals. And I think law-abiding citizens just should not be treated like criminals.”

Update, September 24th: Added statement from Flock.

Can you forget how you feel about Meta?

2026-09-22T11:55:25-04:00

In 2021, as a whistleblower emerged with dramatic allegations that Facebook was harming children, CEO Mark Zuckerberg posted a rebuttal on Facebook. “We care deeply about issues like safety, well-being and mental health. It’s difficult to see coverage that misrepresents our work and our motives,” he wrote in a note to Meta staff, which he reposted publicly. “At the most basic level, I think most of us just don’t recognize the false picture of the company that is being painted.”

Away from the stream of often inane public replies, one reader didn’t like what they saw. Zuckerberg spoke with the unidentified person privately on WhatsApp, according to exhibits later released as part of a series of lawsuits, and the figure urged Zuckerberg — a parent himself — to make child safety even more central in his messaging. “i am not saying facebook is duplicitous and evil,” they wrote. “i am saying the opposite, you have heard the chorus and will PERSONALLY make this issue over adolescents your issue.”

In the privacy of the conversation, Zuckerberg was candid. “I care deeply about this and we’ll continue focusing on being industry-leading with young people,” he responded, “but I’m not going to say it’s my personal main focus when I have a number of other areas I’m more focused on like building the metaverse.”

Zuckerberg later reiterated that “I personally care about this.” But he appeared ambivalent, nudging his conversation partner away from older concerns and toward his vision for the future. “How do you think this fits with our focus on promoting building the metaverse as the next major opportunity?” he asked. “That’s not at odds with safety, but it’s a much different and more innovation-focused message.” Weeks later, Zuckerberg made a huge announcement: The company was changing its name to Meta in honor of its newest, most important priority.

Years after that conversation, online child safety concerns have ballooned — while if anybody’s talking about the Metaverse, it’s likely as a multibillion-dollar money sink where nobody has legs. But the exchange is a microcosm of how Zuckerberg has run Meta, a sprawling online empire worth well over a trillion dollars: by always chasing the next big opportunity. Facebook’s motto was once, famously, “Move fast and break things.” Its new one might be closer to Move fast, break things, move on.

For two decades, Meta’s products have been at the center of what people do online. But Meta now finds itself besieged on all sides. Lawsuits have piled up as regulators turn on social media. A push into smart glasses has triggered a mountain of negative press and backlash — including the pejorative moniker “pervert glasses.” The company remains firmly behind in the AI race. Public perception of Zuckerberg himself has curdled.

Yet Meta is far from finished. It maintains a growing, highly profitable social media business. It’s dominant in AI-powered smart glasses, what many tech leaders consider the next mobile computing platform. It’s defused potential regulatory threats. And its latest platform, a “personal AI agent” dubbed Muse, was released in early September, giving Zuckerberg his next metaverse-like project.

Muse is Zuckerberg’s latest bid to reinvent the company. Early numbers for Muse are promising; an Apptopia estimate this week indicates it’s got over 600,000 daily active users in the US, shooting it to the top of the iOS App Store charts. Yet the question remains: Can a company synonymous with safety and privacy debacles become the AI-powered voice in everyone’s ear?

The glasses gamble

When Zuckerberg took the stage at Facebook Connect to announce that one of the most recognizable names in tech was getting a new identity, he was planting a flag. Facebook, the social media company known for its “big blue app”? That was the past. The future of the internet was the metaverse, an immersive 3D space where digital avatars could socialize, work, and collaborate, built on a virtual reality headset the company acquired in 2014. And after the covid pandemic’s mass shutdowns of physical space, Facebook believed in the metaverse so much that it would rebuild its entire identity around the niche world of virtual and augmented reality.

Nearly five years later, Meta’s Reality Labs division has accumulated over $80 billion in operating losses. It kicked off this year by cutting roughly 10 percent of that division, shuttering three VR studios, discontinuing the metaverse for work, and announcing it’d be shutting down Horizon Worlds, its flagship metaverse experience, on VR headsets. Meta has since decided to keep Horizon Worlds going after public backlash, but it’s clear that Zuckerberg’s colossal gamble on VR hasn’t paid off.

But to call Meta’s investment an abandoned boondoggle isn’t quite accurate either. Meta dominates the nascent but growing smart glasses market, perhaps its biggest success story in recent years. The metaverse may not be the future of computing, but its augmented reality ambitions could still pay off.

Few expected the Ray-Ban Meta glasses to take off when they launched in 2023. After all, Meta’s first attempt, the Ray-Ban Stories in 2021, was a bona fide flop. Ninety percent of owners left the glasses to rot in their drawers. But improvements to cameras in the next-generation hardware helped Meta to sell roughly 2 million units — surprising even Zuckerberg. By the end of 2025, EssilorLuxottica — Meta’s partner in building these devices — had sold an additional 7 million units. All told, Meta and EssilorLuxottica reported that smart glasses sales doubled in Q2 of 2026.

Few expected the Ray-Ban Meta glasses to take off — but their success has surprised even Zuckerberg

Most importantly, Zuckerberg and Meta leadership view this tech as the perfect vehicle for the company’s latest, shiniest interest: its AI investments. So much so that Meta is aggressively trying to brand these as AI glasses.

“I think in the future, if you don’t have glasses that have AI — or some way to interact with AI — I think you’re probably [going to] be at a pretty significant cognitive disadvantage compared to other people and who you’re working with or competing against,” Zuckerberg said in a 2025 earnings call.

Meta is still funneling billions into the AR space without profit, but that’s not necessarily disastrous. “There’s no true way to tell if [Meta] is successful or not. You could measure market share, and they absolutely have that, but that still doesn’t speak to profitability or if the spending to date is justified,” says Jitesh Ubrani, research director at IDC, noting that the runway for XR — the umbrella term for virtual, augmented, and mixed reality — is a notoriously long one.

VR and the metaverse also had long runways that never paid off. But right now, the smart glasses market is building momentum, making it far too early to say whether it’s doomed to the same fate.

Looking at what numbers do exist, Meta is the undisputed king. According to IDC, Meta commands 69.2 percent of the smart glasses and headset market as of Q2 2026. That number jumps up to 76 percent of AI glasses without displays. That said, Ubrani argues that it’s unlikely that any company will see a return on investment in the next five years. It’s possible, he says, that could take as long as 25 years. Absent profit, the next best measure of success is often determined by who can afford to stay in the game long enough — while ignoring demanding investors — to influence the next chapter.

“In the XR market, I’d say that [Meta] is the leader in every segment they play in,” says Anshel Sag, principal analyst at Moor Insights & Strategy. That can be a double-edged sword, though. “They’ve both been a steward of the XR and AR space, but they’ve also limited its potential because they control the platform, the spending, and they decided who were the winners and the losers by investing with their dollars.”

Both Sag and Ubrani also agree that Meta wouldn’t currently be the leader in smart glasses or a major player in AI if not for its early, if flawed, investments in VR.

Meta’s success on this front has kicked off a race among tech companies. Google is pushing its Android XR platform, working with Samsung to launch several “intelligent eyewear” products later this year. Bloomberg has reported that Apple is also working on a pair. Outside the major players, there are dozens of Ray-Ban dupes from smaller companies. Again, Ubrani says it’s difficult to put hard numbers on how much tech companies are pouring into this space — but whatever the cost, the prize is winning what Silicon Valley sees as the next big mobile computing platform.

Ubrani points to the smart glasses supply chain, where “one player who’s willing to invest no matter where the market is going” has massively changed the game. He contrasts that approach with Google, which retreated after criticism of its own glasses over a decade ago. “When people didn’t like [Google Glass], they scaled back significantly, and that in a sense, hurt the entire industry. Now with Meta, we’re seeing that even if Meta’s not making money, the industry is still moving forward.”

“Even if Meta’s not making money, the industry is still moving forward.”

On paper, it might seem Meta’s already won the battle for the next mobile computing platform. At least, if Meta weren’t Meta. While the company gained serious momentum in 2024 and 2025, it’s attracted a public backlash largely of its own making. Meta pushed its glasses as a way to make content for Instagram and other platforms, but failed to account for how bad actors might misuse its products. Online pickup artists and pranksters have started using the devices to harass women and retail workers, resulting in the “pervert glasses” label. The resulting videos are pushed to social media, where they effectively publicize the tech’s most troubling aspects. All this stokes roughly a decade’s worth of privacy concerns stemming from indiscriminate data collection, breaches, and advertising practices. It’s not just that Meta perhaps underestimated how its products could be misused. It already built a reputation for not taking privacy seriously.

In recent months, activists have launched guerrilla campaigns in major cities by “remixing” Meta’s own marketing, painting smart glasses users as privacy threats.

Meta has tried to demonstrate that it cares about privacy. “If people aren’t comfortable with you wearing the glasses, not only do we personally think that’s bad, but we wouldn’t have a business anymore,” Alex Himel, Meta’s VP of wearables, recently told The Verge ahead of a mandatory update that makes it harder to disable the LED light that alerts people that the headset is recording. Instagram head Adam Mosseri has also recently said Meta will take down prank videos captured with the Ray-Ban glasses.

But these have been largely reactive policies. For the public, they look more like attempts to distract from unflattering news that feeds into Meta’s notorious privacy reputation. The New York Times reported that Meta intended to launch facial recognition software during political unrest to throw off privacy advocates. That was soon followed by a Wired report that said Meta had shipped — but not activated — this facial recognition tech to millions of glasses. Rounding out the trifecta was a Financial Times report that Meta is working on “super sensing glasses” that will be able to see what you’re doing 24/7. Andrew Nazdin, director of Glasgow Actions Team, one of the groups behind the guerrilla campaigns, says it would be “foolish for us to let them police themselves instead of calling on society to reject this [tech].”

It’s too soon to say if this backlash has hurt Meta’s smart glasses ambitions. Ubrani says we likely won’t see any impact on sales — if there is one — until the holiday season. Even so, analysts aren’t convinced Meta’s bad reputation is enough to derail its hopes for smart glasses.

“Meta’s done a very poor job of talking about privacy and setting the right tone,” acknowledges Sag. “But it’s funny because if you go [politically] far left or far right, people care more about privacy. In the middle, people care less and just want to have a good experience.”

“If you go [politically] far left or far right, people care more about privacy. In the middle, people care less and just want to have a good experience.”

As for public distrust of Zuckerberg, Sag says his thorny reputation may not sit well with the public, but it’s considered part of Meta’s strength in a niche space like XR.

“Meta would not be where it is today if it weren’t for Mark Zuckerberg … Even if people disagree with him, ultimately what he wants is what gets done,” Sag says. “I think his fear of losing the AI race is what sparked the whole Superintelligence Lab, and now they’re back in the competition. He has blind spots, but he’s able to correct much more quickly than any other CEO founder … every time people have counted him out or questioned his leadership, he’s been able to claw the company back.”

That unilateral doggedness is why Meta seems hell-bent on barreling forward with AI glasses, privacy qualms be damned. A few months ago, it released Meta Glasses, a newer and cheaper version of the Ray-Ban glasses sans Ray-Ban branding. At a launch event, Meta CTO Andrew Bosworth told the press that Meta was contemplating audio-only glasses without cameras. It will likely launch even more new styles, collaborations, and potentially new hardware and privacy protections at Meta Connect 2026 this week.

“Meta missed mobile, they missed VR when it first came out. The metaverse didn’t pan out the way they thought it would. That’s one of the reasons why we’re seeing them so determined and willing to spend to make AI glasses happen,” says Ubrani.

It’s unclear whether Meta can maintain its lead in the XR space. AI wearables are only just beginning to heat up, with players big and small eager to throw their hat in the ring across multiple form factors ranging from pendants to smartwatches. Meta only has glasses — and it’s tying the glasses’ success to the overall success of the company’s AI bets. And on that side of the business, things are looking like a hot mess.

An AI ‘shitshow’

Last summer, Zuckerberg attempted to lure top AI talent to supercharge his company’s lagging AI division. He offered a series of staggering pay packages, running up to $1 billion, to join Meta’s Superintelligence Labs. But more than a year later, the company doesn’t have much to show for it besides two manifestos published by Zuckerberg and the recently released Muse.

Zuckerberg may have inspired confidence among investors and analysts, but sources at or close to the company told The Verge that morale is down, AI initiatives are scattered at best, and the company’s AI strategy is running largely on vibes alone.

“It’s a shitshow,” says one longtime Meta employee, who works on Reality Labs and requested anonymity for fear of retaliation.

Meta dipped its toes into AI around its metaverse restructuring and released a small group of models. But focused on Zuck’s VR ambitions, it missed its competitors investing in upstarts that would soon become AI giants. Its first frontier models were released in 2023 to lukewarm reception, and it wasn’t until the tepid response to Llama 4 in 2025 that the company jumped into the AI race with both feet.

That’s when Zuckerberg kicked off his AI hiring strategy in earnest, even hand-delivering soup to researchers and engineers he wanted to recruit from rivals and other industry heavyweights. In June 2025, Meta invested $14.3 billion in data-labeling company Scale AI and poached its CEO, Alexandr Wang, to lead Meta’s AI efforts. But many AI engineers turned down Zuckerberg’s advances. Meta AI underwent wave after wave of reorganizations, and its superintelligence sub-team, TBD Lab, has reportedly clashed with Meta executives over opposing goals.

TBD Lab has reportedly sought to think longer-term and build best-in-class models to secure the company’s place among its competitors. But executives have reportedly remained focused on more platform-specific ways to drive revenue, like companion chatbots and social media features. Some attempts to integrate AI into platforms like Instagram have drawn backlash — it briefly launched and then suspended a controversial Instagram deepfakes generator, and said it “missed the mark” on an AI tool that provided creepily detailed information about a user’s children.

“Every time a new technology comes that could potentially augment the company’s revenue stream beyond ads, the company kind of goes manic.”

Meta’s most prestigious AI team is a “total mess,” according to Wired reports that describe internal chaos and ever-changing strategy. Zuckerberg reportedly admitted to employees during an internal town hall this summer that progress on AI agents was slower than expected. On a recent Meta earnings call, an analyst told executives he had been fielding questions from investors about Meta’s amount of debt compared to capital, as the company ramped up its spending to try to get ahead in the AI race. Zuckerberg has also been criticized for slashing team sizes to reportedly replace a significant number of employees with AI systems, per Reuters.

The current employee says Meta has a bad habit of jumping from hype cycle to hype cycle: “Every time a new technology comes that could potentially augment the company’s revenue stream beyond ads, the company kind of goes manic in order to invest and to change and to grow, to incorporate this new technology. It happened with crypto. It happened with VR and the metaverse. It’s happening now with AI. I think it stems from a lack of maturity of our founder [and] a lack of maturity of our leadership.”

A former Meta employee, who also requested anonymity, echoes those thoughts: “Like most big tech, there’s just been such a big push to AI-ify everything … but it’s basically a laughingstock internally.”

In a May podcast interview, Wang denied that he and Zuckerberg had been fighting a lot, saying leaders at the company know frontier model development and product integration are both important. But, he added, “like any company … we debate the points deeply.”

But the current employee says Meta’s AI strategy simply isn’t based in the reality of the business or what its customers actually want. “There’s a whole different planet between what the C-suite is talking about and what’s happening on the ground,” they say, later adding, “That tension is literally at the core of the problem at Meta … building AI for the sake of having the biggest building or the biggest fucking dildo versus building a useful product.”

Meta has spent an extraordinary amount on AI, but it’s made few waves in the industry

Zuckerberg’s latest manifesto — in which he lays out his vision of superintelligence that people use “to create and discover extraordinary new things, build new businesses, express new ideas, learn new concepts, and advance our health and quality of life” — didn’t make many waves in the broader AI industry. And so far, the features Meta has added to smart glasses and the products it’s rolled out haven’t broken through compared to the company’s biggest rivals. Given the early usage statistics, Muse could break this pattern — but it’s too early to know.

Noah Smith, senior director of NLP research at the Allen Institute for Artificial Intelligence (Ai2), says the manifesto “reads to me a bit like PR … I also didn’t find anything particularly surprising there, or new.” Instead, Smith found himself asking what is going on behind the scenes to make Zuckerberg publicly state all this now.

Zuckerberg’s manifesto also wasn’t received well internally, according to the current and former Meta employees The Verge spoke with. In particular, the current employee says it’s a typical example of putting the cart before the horse.

“The manifesto did not articulate the biggest question,” he says. “What will AI do for our business as an advertising business? What is the benefit in terms of return on investment on the hundreds of billions of dollars that we’re spending? We don’t even have a product play.” He adds that even after Meta’s massive investments, he wouldn’t rank his own company in the top three of AI players — and that that’s reflected by the fact that engineers at the company don’t typically prioritize using Meta’s own AI models, instead using models from Anthropic, OpenAI, and Chinese AI companies.

He says that after Meta announces a new AI model internally, people will try it out for a little bit and then typically go back to using Anthropic’s Claude models; he compares employees using MetaAI to children goaded into eating their vegetables.

“I think if we actually cared about developing AI well, we would have everyone just use [our] AI and constantly give feedback,” he says. “That would be monumental. But likely we would take a productivity hit in other ways.”

Overall, “morale is in the gutter,” the current employee says. “People do not trust leadership. I get a strong sense that leadership doesn’t even trust Mark in his direction.” They add that employees feel the company is spying on them and forcing AI into everything, and amid restructuring and layoffs, workers are “tired of the pizza parties” and haven’t seen any movement on top internal complaints, like problematic incentive structures, misuse of tokens, and pace of work.

The former employee says that colleagues at the company feel that it’s “like the Metaverse 2.0,” and though Zuckerberg bought some of the best AI talent in the world, “still, they don’t have a thing that anybody wants.”

A lot of people at Meta, the employee says, think it’s an inevitability that the company will drop its AI efforts the same way it sidelined its metaverse ones.

Meta is betting on giving users free, easily packaged access to agentic AI, but its position in the industry is tenuous

“It’s not going to be framed [as him] abandoning anything,” the current employee says. “They were going to turn off Horizon Worlds, and Zuck got wind of it and he said they need to keep it on even though it’s a money-suck … Zuckerberg does not want to admit defeat. But there will be a strategic pivot … away from this manic AI spending to whatever is next, maybe a derivative of AI, maybe more wearable stuff.”

Muse is the next step in Zuckerberg’s plans: a move from straightforward language and image models toward more practically useful agentic AI. It’s betting on giving users free, easily packaged access to the kinds of tools that might previously have required dedicated hosting machines or cost hundreds of dollars a year. But it’s too early to tell if users will trust Meta with the keys to their digital lives. Meanwhile, Meta’s position in the industry’s top tier appears as tenuous as ever.

In response to questions from The Verge, Meta spokesperson Alex Dziedzan said that “at Meta, our philosophy is simple: let our teams deliver. We encourage you to check out Connect to learn more.”

Meta is still trying to convince investors — and the tech industry more broadly — that its relatively nascent AI business has legs. In the meantime, the company is fighting to retain control of its true money machines and the original business proposition: pumping users’ social media feeds, particularly on its flagship Instagram, with whatever keeps them engaged.

The cash cow

Four years ago, Instagram teased what would become perhaps its most reviled change to date: an “immersive” experience where photos and videos took up a user’s full screen and filled it with content from accounts a user didn’t follow. The feature was blatantly ripped from TikTok, its newer, more ruthless competitor, and to many users, the update represented a betrayal of Instagram’s core social elements.

Tati Bruening saw something worth saving in Instagram’s earlier, actually social, peer-to-peer communication platform. The photographer and creative tapped into widespread resentment and frustration, posting a graphic that was simple and to the point: “MAKE INSTAGRAM INSTAGRAM AGAIN,” it read. “Stop trying to be tiktok i just want to see cute photos of my friends. Sincerely, everyone.” Kim Kardashian and Kylie Jenner, titans of attention-based online economy, reposted Bruening’s graphic. An online petition racked up hundreds of thousands of signatures. Eventually, the heat became too much to bear, and Instagram rolled back the changes. But it was only kicking the can down the road; this was the beginning of the end of an era.

Several years on, watching vertical videos of strangers is a cornerstone of social media, including Instagram. And Bruening has a new perspective, having seen platforms rise and fall and upend her job.

“This is going to be the most shocking thing to hear coming out of my mouth, but I think in terms of all of the creator platforms, [Instagram] is going to be the one that makes it out of this era,” Bruening says.

“I think in terms of all of the creator platforms, Meta is going to be the one that makes it out of this era.”

Meta (as Facebook) acquired Instagram in 2012 for approximately $1 billion, and it eclipsed the company’s then-eponymous blue app in relevance years ago. Central to the creator industry, celebrity culture, and lifestyle content, Instagram is a cash cow for Meta: eMarketer estimates it will pull in $52 billion in US ad revenue this year, compared to Facebook’s $48 billion. For the last few years, it’s been playing catch-up not only to TikTok but to a whole internet where people spend less time connecting with friends and relatives and more on an algorithmically sorted recommendation feed.

Yet along with its rival YouTube, Bruening says, Instagram remains one of the few social media platforms that’s been somewhat stable. Snapchat no longer serves creators, and TikTok is too filled with slop and TikTok Shop content. Meta may have turned its photo app into a vertical video slot machine, but at least it rewards the creators who obey — if you make shortform videos, Instagram will show them to someone.

Creators and brands still see Instagram as the prestige, more polished platform where most advertising spend goes, says Lia Haberman, an influencer marketing expert who writes ICYMI, a social media newsletter. Whatever lead TikTok had in shortform content a few years ago has narrowed or even closed. The threat of TikTok disappearing from the US made advertisers nervous, Haberman says; creators started to think about backup plans and diversifying their platforms and online presence. When brands ask creators for deliverables — the assets and content that they want an influencer to post — Instagram Reels and Stories are typically first on the list, in Haberman’s experience.

Instagram is Meta’s cultural crown jewel, but the company is trying to diversify its revenue through other apps, too. Meta has tried to pull more creators onto Facebook with cash incentives. The company’s massive messaging platform WhatsApp remains popular, especially outside the US; executives said subscriptions and paid messaging drove record non-advertising revenue. Threads, the X competitor, reached 500 million monthly users this summer, but it still feels a bit like the black sheep — in earnings reports, Meta’s “family of apps” is defined as “Facebook, Instagram, Messenger, WhatsApp, and other services.”

The company has introduced a slew of other apps, like standalone platforms for Facebook groups and Marketplace, with leadership saying AI is speeding up development.

“Social media is saturated in many ways. The hyper growth days for Meta are over.”

But despite Zuckerberg’s soliloquies on an AI-powered future, Meta is at its core in the business of serving ads, and that business is still booming: The company reported in July that its revenue from advertising grew 27 percent year over year. Some projections show Meta’s ads business surpassing Google’s by the end of 2026.

The bigger problem is the people it shows those ads to, and the content it puts those ads in between. The average social media user is less interested than they once were in posting, and Meta’s access to a new generation of users is under threat.

“There is sort of a tension brewing where we’re seeing this double-digit growth in their ad spend, but we are not seeing double-digit growth when it comes to overall usage [in the US],” Minda Smiley, senior analyst at eMarketer, says of advertisers’ spending patterns. “I think that just really speaks to the fact that social media is saturated in many ways. The hyper growth days for Meta are over.”

People in general are now using social media to consume content rather than share it; a Morning Consult poll found that a third of respondents say they post less on social media compared to last year. What’s more, entire generations of young people are logging off or forcibly being logged off.

The past few years have seen bipartisan, multinational, and cross-generational support for either limiting minors’ access to social media or banning them altogether. When kids can get online, Gen Z and Gen Alpha don’t want to express themselves fully on the social media sites their parents or relatives are on, and what’s more, they want in-person experiences, Haberman says. They want to touch grass.

Katie Harbath, a former Meta public policy director who now runs tech policy consultancy Anchor Change, says that there’s a small point of relief for Meta here, though: Kids aren’t how they’re making their money. Data compiled by research platform BusinessStats primarily from Facebook’s Ads Manager found that about 3.2 percent of Facebook’s global audience was between the ages of 13 and 17. On Instagram, using a similar dataset, that share was slightly higher, but still just 5.8 percent of the app’s global audience, according to BusinessStats.

“If all of a sudden they can’t have kids under 18 on, it’s not great, but it’s also not the vast majority of the people that are on the platform,” Harbath says.

But Meta will eventually need those users — it can’t dine off the social media habits of older generations forever. What happens to Meta if its user base peters out, and advertisers follow? A reckoning could be coming in two to three years, Haberman says, if these trends continue as Gen Alpha comes of age.

Teen users are a tiny part of Meta’s business — but losing the next generation is something it can’t afford

“I’m not sure that we’ve really fully grasped yet how much of a big deal that this could be once audiences stop showing up or audiences age out,” Haberman says. “Essentially, it’s going to become kind of a graveyard for millennials, Gen X, and boomers.” The implications are manifold — for the social platforms themselves, naturally, but also for the advertisers, brands, and influencers who’ve taken for granted that social media is an effective way to reach people.

The elephant in the room is AI — Meta’s costly buildouts and investments, of course, but also the question of how to handle AI content on the platform, and how much of it users actually want. On its second quarter earnings call, Zuckerberg said the company would soon be able to provide a “nearly infinite universe of personalized content” to users through its image and video generation tools, to supplement content from influencers, creators, and users’ friends and family. AI content is cheap, fast, and replaceable, made to be chewed up and spit out as long as the algorithms will allow it to be served up. This is what creatives like Bruening are competing against.

“You would have to become an AI robot in order to keep up with a lot of the metrics and standards for what [Meta’s] best practices are,” Bruening says. “A human artist is not going to have the same level of output as an AI machine that’s going to appease the algorithm.” How far can Meta push the human creators before they decide they no longer want to compete with their automated counterparts?

Much like Meta, creators are constantly chasing the next big thing, trying to weather shifting societal expectations and tastes. For individual creators, a failure to adapt can be career-ending; Meta now finds itself in similar precarious footing. But concerns about how people use social media aren’t just a matter of business. They’re the object of serious inquiry for lawmakers across the world — all bent on reining in what they see as one of the tech industry’s most dangerous players.

Over the past decade, Meta has become one of the biggest punching bags in politics — and its CEO the face of a callous, out-of-touch, and exploitative “Big Tech.” Zuckerberg, Mosseri, and other executives have been dragged before congressional committees and juries and berated for a perceived carelessness toward kids’ safety and disinformation, a ruthless crushing of competition, and even complicity in lynching and genocide. But through each PR crisis, Meta has somehow still found its footing, and in many cases, dodged the worst-case outcomes.

Meta grew to massive scale and influence in part through two key acquisitions: its purchases of Instagram in 2012 and WhatsApp in 2014. But by 2020, those deals became the focal point of a major antitrust lawsuit seeking to break up the company. Since then, Meta’s acquisitions and competitive actions have faced intense scrutiny in the US and around the world, and its dealmaking leveled off. In 2020, it closed 10 acquisitions worth a total of $1.37 billion, according to data compiled by PitchBook; as it fought monopolization claims, its deal count and value dropped each year through 2024.

But for the most part, the company has weathered regulatory pressure, and Meta’s dealmaking picked up again in 2025. Meta won against the FTC’s anti-monopoly claims, and it was allowed to buy the virtual reality fitness startup Within after the FTC took it to trial (a few years later, it announced it would stop making content for Within’s flagship game, only to later spin it off).

Through political crisis after political crisis, Meta has dodged the worst-case outcomes

Long-standing fears about Meta’s data collection efforts also reached a boiling point in 2018, when reporters uncovered evidence that London elections consultancy Cambridge Analytica had harvested millions of Facebook profiles to analyze and target users with pro-Trump messaging in 2016. That led to a flood of calls among users to #DeleteFacebook, even from Brian Acton, cofounder of WhatsApp.

But Zuckerberg later told The New York Times that the movement hadn’t resulted in any “meaningful” defection from the platform. In Europe, Meta was ordered to make changes impacting its ad-targeting practices after a $1.3 billion fine, but in the US lawmakers are still unable to agree on an industry-wide data privacy law.

While President Donald Trump once threatened to throw Zuckerberg in prison for life, the CEO seems to have successfully gotten on the Trump administration’s good side after ditching the fact-checking program the president had criticized, disbanding its diversity, equity, and inclusion (DEI) team, and agreeing to a $25 million settlement over Trump’s lawsuit for suspending his account. Meanwhile, Meta’s apps benefited from the brief ban of TikTok in the US, though Trump cut that period short through repeated enforcement delays. Meta’s courting of the Trump administration may leave it ripe for scrutiny under a Democratic administration, but meaningful consequences could remain years away, if they come at all.

Most recently, Meta’s most pressing policy challenge has been child safety. For the first half of the year, it appeared to be caught on its back foot. Two juries issued decisive blows to the company, resulting in close to a billion dollars in fines across an individual plaintiff’s case in Los Angeles and a case brought by New Mexico’s attorney general. But the New Mexico judge later declined to impose the most disruptive structural changes the AG had asked for.

On the surface, a landmark settlement with 47 other state AGs in August looked even worse for Meta. The settlement included significant changes: new age assurance benchmarks, a two-hour daily usage limit (with exceptions for messaging and longform content), and default restrictions on nighttime and school hour notifications for teens.

But it also presented Meta an opportunity to turn attention onto its rivals. The company called on TikTok, YouTube, and Snap to adopt similar limits, and it tied part of its settlement payout to their actions. This helped Meta twice over: It could put a target on its competitors’ backs while mitigating the potential negative business impact of its own deal.

New child safety regulations could end up helping Meta as well. eMarketer forecasts that Snap and TikTok would feel more of a hit than Meta if a nationwide classroom cellphone ban, another popular proposal, went into effect. TikTok would lose out on more than $1.25 billion in ad revenue this year, the firm estimates, and Snap’s ad revenue would be especially vulnerable because of the high concentration of teens on the platform who use the app throughout the day. A hit on Meta’s revenue, meanwhile, would “hardly register,” eMarketer says, thanks to its scale.

Meta could lose out under child safety regulations — but its competitors might lose more

Even so, Meta’s legal battles over kids’ safety aren’t over. Meta and other tech companies still face thousands of lawsuits from individual plaintiffs and school districts, likely culminating in a blanket settlement. The scope of cases and potential penalties is so large that it’s difficult to understand exactly how much risk they carry. It would take quite a lot for Meta to feel the sting, but the billions in settlements and court orders are beginning to add up.

Meta has already invested huge sums into fighting these cases. In the second quarter alone, it took a $2.4 billion charge over litigation expenses. Prior to Meta’s recent settlement with state AGs, Peter Chapman, deputy director of the Knight-Georgetown Institute, said that “suggests to me that they’re playing a long game” around the First Amendment and online liability shield Section 230.

Following the settlement news, Chapman says he still sees it as early days amid the mountain of litigation before Meta and other tech companies, but “the long game certainly has changed.” Now it seems to include targeting industry competitors to get on board with the terms Meta already accepted, while a fight to preserve its protections under Section 230 will likely continue through appeals of the earlier cases Meta lost.

Meta’s settlement may bring some solace to investors who were unsure how to price in its legal risk. After the news of the deal, its stock price shot up. “The state AGs got an okay deal. I don’t think they knocked it out of the park,” says Paul Gallant, a tech policy analyst at investment bank TD Cowen. As a result, Meta shareholders should have “confidence that the company knows how to manage the government.”

The $17 billion settlement represents an “immaterial” amount for Meta, Gallant says. In the last quarterly earnings report, Meta said it made more than $60 billion. But additional cases still present some risk, and large fines may be more of a problem today than when the European Union levied them in the past. “The companies are spending so much on AI infrastructure that cashflow is a bigger focus of Meta’s stockholders, especially if the damage awards get into the tens of billions or more,” Gallant says. And unlike the tactics at its disposal to fight off impending legislation, Gallant adds, “of all of the features of the US governmental system, the least controllable by Meta is juries.”

“These massive trials are happening against this company, and then what is Mark Zuckerberg doing? Putting out an AI manifesto.”

State and local governments also appear ready to take on tech while federal efforts stall, including in Meta’s latest area of focus, AI. Many are looking at how they can rein in AI and deal with the energy demands it creates on their electrical grids. Data centers have become a flashpoint across the country, and a nationwide push for a slowdown could be particularly meaningful for Meta and other businesses in the AI race. “That’s the thing that could really not just slow them, but slow everybody down — having enough compute,” says Harbath. “We talk so much about what happens at the federal level, but I actually think the biggest risk to Meta and these tech companies right now is at the hyper-local level.”

Zuckerberg has become the face of many of these battles, testifying for hours before policymakers and juries, and even standing up in Congress to apologize to parents who lost children after battles with online harms. The public testimonies have largely not seemed to endear him any further to his critics.

“It’s almost like this bizarre disconnect where these massive trials are happening against this company, and then what is Mark Zuckerberg doing? Putting out an AI manifesto,” Smiley, the analyst at eMarketer, says. “It’s sort of like, ‘What planet are you on?’ It just feels like there’s just this big disconnect.”

Whether it will matter is another unresolved question.

Despite the litany of PR problems, user frustrations, legal troubles, and internal personnel conflict, Meta still feels like an immovable force. It has managed to sidestep, acquire, and pivot its way out of scandals and blowups that would have sunk most other companies. It’s spending lavishly on its new, AI-powered future.

But that future remains far from certain. Smart glasses are unproven, its AI labs are behind newer upstarts, the legal landscape is hostile, and it may be hitting its ceiling in finding new users.

What does seem certain is that its CEO will never stop chasing the next big thing. “Zuckerberg has this kind of monkey on his back,” says the current Meta employee. “He’s willing to experiment with the company because he controls the company.”

Virginia governor creates an AI task force and moves to restrain data centers

2026-09-18T15:09:19-04:00

Virginia Gov. Abigail Spanberger ordered the state government to take steps that could empower local communities to have a larger say in data center development and slow down approvals in a state that is already home to the data center capital of the world.

Executive Order 22 bans executive branch officials from signing nondisclosure agreements for data center projects, requires expedited noise regulations, and requires a review of backup-generation operations used by data centers, among other requirements. The order also establishes an AI task force responsible for evaluating how the state government can address risks to Virginians like workforce displacement and data privacy issues, and determining how existing law can be applied to AI harms.

The order represents the latest signal that states are taking data center and AI policy into their own hands as Congress and President Donald Trump have been slow or resistant to sweeping changes. Also on Friday, California Gov. Gavin Newsom issued an executive order convening experts to tackle AI safety issues, and earlier, New York Gov. Kathy Hochul and Texas Gov. Greg Abbott have both taken executive action to rein in rapid data center development.

In addition to Virginia’s new executive order, Spanberger unveiled a new “Data Center Accountability Framework” for the state, which lays out additional priorities. These include eliminating “by-right approval,” like the kind that existed in Loudoun County for many years allowing developers to build on certain land without additional approvals, removing some state subsidies, establishing environmental guardrails, and safeguarding residents against increased energy prices that are associated with data centers.

A press release for the actions includes many statements from environmental groups praising Spanberger’s approach, but others reportedly wish she’d gone even further. Piedmont Environmental Council (PEC), for example, is a nonprofit that has advocated for a moratorium on new development to deal with existing concerns, and PEC president Chris Miller said in a statement that “It largely addresses standards on new data centers and does not deal with what’s already built and in the pipeline, along with the ongoing impacts piling up in communities now.”