Garrison Lovely | The Verge

Hollywood is coming out in force for California’s AI safety bill

2024-09-24T17:06:53-04:00

Governor Gavin Newsom gave his thoughts on SB 1047 for the first time at Salesforce’s Dreamforce event last week. | Photo by Justin Sullivan / Getty Images

Hollywood is squaring off against Silicon Valley in the battle over SB 1047, California’s first-of-its-kind AI safety bill. Amid doubts about whether Governor Gavin Newsom will sign the legislation, a wave of star-studded endorsements mark the first organized celebrity effort to advance AI regulations beyond the direct interests of the entertainment industry.

On Tuesday, over 125 big Hollywood names published an open letter urging Newsom to sign the AI safety bill. Signatures include Ava DuVernay, Jane Fonda, J.J. Abrams, Shonda Rhimes, Alec Baldwin, Pedro Pascal, Jessica Chastain, Adam McKay, and Ron Perlman. “We fully believe in the dazzling potential of AI to be used for good. But we must also be realistic about the risks,” the letter reads. In a sign of genuine enthusiasm, the letter was written by one of the signatories, according to a person in contact with the celebrities.

SB 1047 is the US’s most significant AI safety legislation to date, and Newsom’s signature would break the precedent of letting the industry police the development and deployment of its most powerful models via voluntary commitments. The core of the bill mandates that the largest AI developers implement safeguards of their own choosing to reduce the chance their model causes or enables a disaster, like a severe cyberattack or pandemic. It would apply to any covered AI company doing business in California, which is home to the top five generative AI companies and the world’s fifth-largest economy. This makes it a de facto national regulation in a country that has trailed the EU, China, and the UK in efforts to regulate AI.

The bill passed California’s legislature in August, and Newsom has until September 30th to sign or veto it. So far, it’s not clear what he’ll do. The governor made his first direct comments about SB 1047 in an interview with billionaire Salesforce founder Marc Benioff last week. While he told the Los Angeles Times he had not made a final decision right before the interview, he suggested to Benioff he had reservations about the bill’s potential impact on the state’s competitiveness.

It wasn’t initially clear how much SAG-AFTRA would go to bat for SB 1047

At the same time, Newsom recently signed two AI bills that directly appeal to SAG-AFTRA at its Los Angeles headquarters, flanked by leaders of the powerful union of screen actors and performers. The labor group was a driving force behind the bills, which regulate the use of digital replicas and build on concessions won in the 2023 actors strike. The union is also one of the most prominent supporters of SB 1047.

The celebrity letter signatories, many of whom are SAG-AFTRA members, write that they are “genuinely grateful” for Newsom’s signature on the two digital replica bills. But the AI safety bill, they write, “is not about protecting artists — it’s about protecting everyone.”

SB 1047 has rallied both support and intense criticism. It’s opposed by most of the AI industry, former House Speaker Nancy Pelosi, San Francisco Mayor London Breed, and eight congressional Democrats from California. Powerful supporters include the Service Employees International Union (SEIU), the Latino Community Foundation, Elon Musk, and — in a letter first reported by The Verge — SAG-AFTRA.

Newsom’s decision may come down to raw power politics — something each side recognizes as they muster support in the bill’s final days. As he contemplates his political future, the governor faces a difficult choice between the two industries his state is best known for: tech and entertainment.

It wasn’t initially clear how much SAG-AFTRA would go to bat for SB 1047. Performing artists have been directly impacted by the precipitous rise of generative AI, from nonconsensual deepfake pornography of celebrities like Taylor Swift to the threat of digital replicas killing acting jobs. In an earlier flashpoint between Hollywood and the AI industry, OpenAI was accused of ripping off Scarlett Johansson’s voice for use in ChatGPT. But unlike the two bills Newsom signed, SB 1047 doesn’t directly address these issues. It’s aimed at far more catastrophic threats.

As the deadline approaches, however, the union’s support is shaping up to be much more than a marriage of convenience with other supporters of big tech regulation. Tuesday’s letter includes SAG-AFTRA leaders like president Fran Drescher, national executive director and chief negotiator Duncan Crabtree-Ireland, secretary-treasurer Joely Fisher, and 12 national board members, including Sean Astin and Rosie O’Donnell.

“We’re the canaries in the coal mine”

Fisher told The Verge that in addition to the protections Newsom signed into law, “our community also cares deeply about safe and responsible AI development, so many of us have come together to urge the Governor to sign this one for humanity and his legacy. Putting guardrails on Big Tech won’t stifle innovation.”

On Tuesday afternoon, Astin published a detailed personal letter he just sent to Newsom, writing that SB 1047 is still required “because it is the only bill that seeks to regulate the largest of the large language models and computer clusters used by giant tech companies.” The Lord of the Rings star also mentions his recently earned master’s degree in public administration.

Prior to Tuesday’s letter, at least 11 SAG-AFTRA members posted on social media in support of SB 1047. Stars like Joseph Gordon-Levitt, Mark Ruffalo, Alyssa Milano, Astin, and Piper Perabo published videos and posts urging Newsom to sign the bill.

Comedian Adam Conover said in his video, “I’m a big critic of AI and a lot of what you hear about the robot uprising is fluff and BS and marketing, but the billionaires who run this industry really are causing real risks they are not doing enough to prevent.”

SAG-AFTRA board member Jason Winston George, a TV star on Grey’s Anatomy, told The Verge that he’s “a huge Gavin Newsom fan” and is “incredibly thankful” the governor signed two union-backed bills. But he doesn’t want Newsom to stop there. George is all in on SB 1047, too.

George brought up Christopher Nolan’s Oppenheimer, which won Best Picture for its portrayal of the “father of the atomic bomb,” likening the anguished physicist to Geoffrey Hinton, the pioneer of deep learning who famously resigned from Google to speak freely about the risks of the technology he shaped. Hinton, the most cited living scientist, has also written in favor of SB 1047.

With the 2023 actors strike, SAG-AFTRA has been at the forefront of efforts to regulate AI in the US. The use of AI tools in Hollywood was perhaps the biggest obstacle to resolving the 118-day actors strike. “We’re the canaries in the coal mine, and we became very acutely aware of how much [AI] is going to change everything, and so I think that’s why our membership is so vocal,” said George, who was part of the union’s negotiating team. “I’m not afraid of AI, I just want to make sure that there are guardrails on it,” he added.

The celebrity support for SB 1047 is another sign of the growing rift between elites in Hollywood and Silicon Valley. A number of high-profile tech billionaires have recently pivoted to the right, whereas the biggest stars have remained solidly blue and endorsed Kamala Harris.

The union is no stranger to using legislation to advance its interests. It currently lists 18 federal and state AI bills besides SB 1047 on the group’s 2024 legislation tracker, primarily covering entertainment-related concerns like AI replicas, deepfakes, and copyright protection. But its support for SB 1047 marks the first time the group has thrown its weight behind an AI bill not directly related to its industry.

Will it make a difference? Nobody seems sure. But Tuesday’s letter includes some of the governor’s most well-known allies.

Hollywood came out in force to back Newsom in his 2021 recall race. And now, some of those prominent supporters have signed on to the SB 1047 letter, like Mark Hamill, Alyssa Milano, Shonda Rhimes, and J.J. Abrams, who donated $32,400 to Newsom’s 2022 campaign. Additionally, Jane Fonda fought oil drilling with Newsom, and Ava DuVernay was inducted by the governor into the California Hall of Fame in 2022.

In 2024, California passed more than 40 AI bills addressing things like self-driving trucks, election misinformation, and AI-generated child pornography. However, none of the other AI bills on Newsom’s desk have provoked as much industry pushback as SB 1047.

“This is a space where we dominate and I want to maintain our dominance”

The governor has been willing to stand up to powerful industries before, like fossil fuels and fast food, but he hasn’t shown that same backbone with big tech. In August, Newsom orchestrated a backroom deal to kill proposals to fund local journalism through taxes on tech platforms like Google and Meta. Instead, Google and California will each contribute money to fund local journalism and an AI accelerator. Leaders of multiple journalist unions called the agreement “disastrous.” (One of the proposals, which was also backed by SAG-AFTRA, did face some good-faith criticism for how it planned to raise and distribute funds.)

At Dreamforce, Newsom spoke at length about SB 1047. While he said he felt a “deep sense of responsibility to address some of those more extreme concerns that I think many of us have,” he seemed to be leaning toward a veto. Newsom worried about the “outsized impact” and “chilling effect, particularly in the open source community” he thinks the bill could have.

“This is a space where we dominate and I want to maintain our dominance,” he said. “What are demonstrable risks in AI and what are the hypothetical risks? I can’t solve for everything — what can we solve for?”

The person in touch with celebrity supporters, who wished to remain anonymous due to the controversy around the bill, told The Verge that some feel “Newsom has played them” by signing the AI replica bills to great fanfare. Some of them “think it was intentional misdirection to make people think that he had signed all the AI bills that they care about, and then, in fact, veto what I’ve heard referred to several times as the ‘big one,’” the person said. “I think that these are very, very smart people. Many of them have read the bill all the way through, have developed opinions on specific sections and provisions in the bill.”

They emphasized that not everyone felt this way, however, and that many “have faith that [Newsom will] do the right thing.”

The Hollywood letter warns that AI’s risks won’t stay hypothetical for long. “Grave threats from AI used to be the stuff of science fiction, but not anymore,” it says. It cites an open letter supporting SB 1047 signed by over 110 current and former employees of top AI companies, who write, “We believe that the most powerful AI models may soon pose severe risks, such as expanded access to biological weapons and cyberattacks on critical infrastructure.”

Its signatories paint SB 1047 as a struggle against a powerful, reckless, and loosely regulated industry. “Advanced artificial intelligence is being deployed by massive for-profit tech companies with very little government oversight,” says signatory Adam McKay. “I’m calling on Governor Newsom to demonstrate that he represents the public interest and won’t cave to Big Tech.”

The Hollywood letter makes a personal appeal. “We are your supporters. We voted for you,” it says. “We want to continue to believe that you are a leader who will stand up for everyone’s wellbeing, not just for a few Silicon Valley giants.”

In this high-stakes drama between Hollywood and Silicon Valley, the final act belongs to Governor Newsom.

One of California’s most influential unions weighs in on AI safety bill

2024-09-11T17:00:00-04:00

As California Governor Gavin Newsom weighs signing or vetoing the fiercely contested AI safety bill SB 1047, SAG-AFTRA and two women’s groups are pushing him to approve it — adding even more voices to an already frenzied debate. The performers union, the National Organization for Women (NOW), and Fund Her have each sent letters to Newsom, all of which have been obtained by The Verge and are being published here for the first time.

The letters from SAG-AFTRA, NOW, and Fund Her highlight concerns about AI’s potential to cause catastrophic harm if the technology is left unregulated. SAG-AFTRA outlines SB 1047’s mandate for developers to test for and safeguard against AI-enabled disasters, like cyberattacks on critical infrastructure or bioweapon development. NOW and Fund Her cite grave warnings from people at the forefront of AI and discuss the technology’s potentially disproportionate impacts on vulnerable groups.

SAG-AFTRA posted a call for support yesterday on X from its 160,000 members, which include stars like Scarlett Johansson and Tom Hanks. NOW, the largest feminist organization in the US with around 500,000 members, said it was motivated by expert claims “about how dangerous this incredible technology can be if it is not developed and deployed responsibly.” Fund Her, a PAC that has helped elect 12 progressive women to prominent positions in California in 2022, writes of the “race to develop the first independent thinking AI,” at which point “it will be too late to impose safety guardrails.”

SAG-AFTRA and NOW represent the latest power players to weigh in on the California bill, which has become the object of exceptional national interest and scrambled conventional partisan boundaries.

SB 1047, authored by state Senator Scott Wiener, would be the most significant AI safety law in the US. It establishes civil liability for developers of next-generation AI models like ChatGPT if they cause disasters without implementing appropriate safeguards. The bill also includes whistleblower protections for employees of AI companies, garnering support from OpenAI whistleblowers Daniel Kokotajlo and William Saunders.

“The AI safety standards set by California will change the world”

NOW writes in its letter that “the AI safety standards set by California will change the world,” a view echoed by bill cosponsor Dan Hendrycks, director of the Center for AI Safety. Hendrycks tells The Verge that SB 1047 could be Newsom’s “Pat Brown moment,” referring to California’s then-governor signing a groundbreaking auto tailpipe emissions law in 1966. He quotes what’s since become known as the California Effect: “where California leads on important regulation, the rest of the country follows.”

Having passed both houses of the state legislature with strong majorities in late August, the bill now awaits Governor Newsom’s decision, due by September 30th. The governor’s office said it doesn’t “typically comment on pending legislation. This measure will be evaluated on its merits.”

This comment notwithstanding, the fate of SB 1047 may come down to a political calculation — a reality each side appears to recognize as they marshal support in the bill’s final hours.

The odd political coalitions that have emerged in the fight over SB 1047 augur a topsy-turvy future for AI policy battles. Billionaire Elon Musk aligns with social justice groups and labor unions in supporting the bill, while former House Speaker Nancy Pelosi, progressive House Congressman Ro Khanna, Trump-supporting venture capitalist Marc Andreessen, and AI “godmother” Fei-Fei Li are all opposed.

AI is the rare issue that hasn’t yet sorted into clear partisan camps. As the technology grows in importance, the debate over how to govern it is likely to grow in intensity and may continue to scramble the usual allegiances.

These recent letters join support for the bill from organizations like the nearly 2-million-strong SEIU and the Latino Community Foundation.

SAG-AFTRA has been home to some of the most organic anti-AI sentiment. Many screen actors see generative AI as an existential threat to their livelihoods. The use of the technology was a major sticking point in the 2023 actors strike, which resulted in a requirement that studios get informed consent from performers before creating digital replicas of them (actors must also be compensated for their use).

“SAG-AFTRA knows all too well the potential dangers that AI poses”

The union’s letter writes that “SAG-AFTRA knows all too well the potential dangers that AI poses,” citing problems experienced by its members in the form of nonconsensual deepfake pornography and theft of performers’ likenesses. It concludes that “policymakers have a responsibility to step in and protect our members and the public. SB 1047 is a measured first step to get us there.”

In a phone interview, organization president Christian Nunes said NOW got involved because the group is worried about how unregulated AI can affect women. She and NOW have previously supported efforts to ban nonconsensual deepfakes.

In the NOW letter, Nunes writes that the dangers warned of by AI experts “would disproportionately fall on vulnerable groups, including women.” She highlights Newsom’s “courageous support for us in the face of intense lobbying pressure” on reproductive rights, equal pay, and paid family leave, and that this support “is one of the reasons why women have voted for [him] time and time again.”

While SB 1047 isn’t explicitly designed to address these groups’ more central concerns, the organizations seem to see strategic value in joining the coalition behind it. Nunes told The Verge she views the bill as part of a broader project to hold Big Tech accountable.

This support for SB 1047 complements other pending AI legislation that more directly addresses these groups’ specific issues. For instance, the federal NO FAKES Act aims to combat deepfakes, while another AI bill on Newsom’s desk, endorsed by SAG-AFTRA, would regulate the use of digital replicas. By backing SB 1047 alongside these more targeted initiatives, these organizations appear to be taking a comprehensive approach to AI governance.

The NOW and Fund Her letters both draw parallels between unregulated AI and the history of social media. Fund Her founder and president Valerie McGinty writes to The Verge, “We have seen the incredible harm social media has imposed on our children and how difficult it is to reverse it. We won’t be stuck playing catch up again if Governor Newsom signs SB 1047 into law.”

It’s unclear if the letters will be enough for the bill to overcome the powerful forces arrayed against it. While Wiener and other advocates describe the regulation as “light-touch” and “common sense,” the industry is, by and large, freaking out.

The US currently relies almost entirely on self-regulation and nonbinding voluntary commitments to govern AI, and the industry would like to keep it that way. As the first US AI safety regulation with teeth, SB 1047 would set a powerful precedent, which is a likely motivation behind both these letters and the vigorous industry opposition.

Google, Meta, and OpenAI took the unusual step of writing their own letters opposing the bill. Resistance from AI investors has been even stiffer, with the prestigious startup incubator Y Combinator (YC) and the venture fund Andreessen Horowitz (a16z) leading a full-court press to kill SB 1047. These and other prominent opponents warn that the bill could prompt an exodus from California, cede the US lead in AI to China, and devastate the open source community.

Naturally, supporters dispute each of these arguments. In a July letter addressing YC and a16z’s claims about the bill, Wiener points out that SB 1047 would apply to any covered AI company doing business in California, the world’s AI hub and fifth-largest economy. Dario Amodei, CEO of leading AI company and eventual de facto SB 1047 supporter Anthropic, called the threat to leave “just theater” (it has nonetheless also been invoked by OpenAI, Meta, and Google).

Nancy Pelosi called the bill “well-intentioned but ill informed”

In her statement opposing the bill, Pelosi called it “well-intentioned but ill informed.” In a phone interview, Wiener said, “I have enormous respect for the Speaker Emerita. She is the GOAT,” but went on to call Pelosi’s statement “unfortunate” and noted that “some of the top machine learning pioneers on the planet support the bill,” citing endorsements from deep learning “godfathers” Geoffrey Hinton and Yoshua Bengio. Wiener also highlights a supportive open letter published Monday from over 100 employees and alumni of the leading AI companies.

For evaluating SB 1047 on its merits, the most convincing letter might be one published by Anthropic, which broke from its peers to write that the revised legislation’s “benefits likely outweigh its costs.” This letter followed a round of amendments made directly in response to the company’s prior complaints. Anthropic’s Claude family of chatbots leads the world on some metrics, and the company will likely be one of the handful of AI developers directly covered by the law in the near future.

With key congressional leaders promising to obstruct substantive federal AI regulations and opposing SB 1047, California may go it alone, as it already has on net neutrality and data privacy. As NOW’s Nunes writes, the “AI safety standards set by California will change the world,” giving Governor Newsom a chance to make history and model “balanced AI leadership.”

Fund Her’s McGinty summed up the supporters’ stance in an email to The Verge: “We should listen to these experts more interested in our wellbeing than the Big Tech executives skimping on AI safety.”

As the September 30th deadline approaches, all eyes are on Governor Newsom to see how he’ll shape the future of AI governance in California and beyond. “My experience with Gavin Newsom is — agree or disagree — he makes thoughtful decisions based on what he thinks is best for the state,” says Wiener. “I’ve always appreciated that about him.”

Correction: The article initially cited deep learning “godfather” Yann LeCun as a supporter of SB 1047. LeCun is opposed to the bill. We regret the error.