The Future of AI Regulation: SB 1047

Author: V. Iyengar, Megaphone Editor-in-chief

On August 28th, 2024, the California state assembly passed a landmark AI regulation bill, SB1047, which will now be sent to the Governor’s desk, where he has until September 30th of this year to sign it.

What is SB1047?

SB1047 was proposed earlier this year by CA state senator Scott Wiener and hopes to regulate the advanced AI models that are being created by companies doing business in California. The bill would require developers of these models to, among other things, make sure that before using or releasing an AI model, it will not cause nor enable harm to neither individuals nor society, and to implement safeguards and take other steps to minimize the amount of harm their models could cause.  Additionally, it creates whistleblower protections for tech workers.

However, to protect California’s startup culture, this bill will only regulate models that are trained at costs that exceed $100 million.

Strong Opposition (and Support)

Despite being seemingly tame in terms of the reforms this bill would create, it has garnered strong opposition from some of the world’s leading voices in AI. AI companies such as OpenAI, Meta, Google and Y Combinator have come out against SB1047 and some, such as venture capital firm Andreessen Horowitz, have spent millions of dollars lobbying to get the bill watered down or killed entirely. Additionally, some academic researchers, along with Fei-Fei Li, who is rumored to be working on her own AI startup at the moment, known as the the “godmother of AI”, have come out in criticism of the bill, expressing worries that it will shackle innovation and threaten open source AI models.

Despite controversy surrounding the bill, it is far more watered down than many think, due to the several rounds of edits and amendments that the bill has been put through. In fact, one prominent critic of the bill claimed that AI developers could be guilty of a felony, regardless of whether they were involved in a harmful incident caused by the model they developed or not, while the bill only had provisions that would hold the developer criminally liable if that developer knowingly lied while under oath, and, ironically even this provision has been removed. 

Interestingly, despite its prominent opponents, it has also attracted support from some influential people in the AI space. Notably, Elon Musk has surprisingly voiced his support for the bill, writing on X, “This is a tough call and will make some people upset, but, all things considered, I think California should probably pass the SB 1047 AI safety bill…For over 20 years, I have been an advocate for AI regulation, just as we regulate any product/technology that is a potential risk.” Even more surprising is that Musk’s own AI company, xAI will be subject to the regulations in SB1047. 

Perhaps more significantly, the bill also received endorsements from deep learning “godfathers” Geoffrey Hinton and Yoshua Bengio, who referred to the provisions in the bill as a “bare minimum for effective regulation.”

A number of nonprofit organizations have also come out in support of SB1047. In an open letter, the Screen Actors Guild – American Federation of Television and Radio Artists (SAG-AFTRA), expressed that they felt 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.” The Screen Actors Guild has been vocal about the fact they they perceive generative AI to be a threat to their livelihoods, an opinion they expressed via the writer’s strike that occurred earlier this year.

NOW, the National Organization of Women, has also come out in support of SB1047 recently, and has been and continues to be an enduring and vocal supporter of regulating AI to protect the public, especially women, who make up the vast majority of deep-fake victims, from AI-related crimes, especially deep fakes and other forms of content modification.

While SB1047 doesn’t directly protect writers from being replaced by generative AI, nor does it crack down on deep fake/content modification-related crimes, these organizations view the bill as pushing forward the movement to hold big tech accountable.

Why is SB1047 Important?

In general, at the federal level, lawmakers have been hesitating to regulate AI, and as of right now, there exist no comprehensive federal legislation/regulations that regulate the use and development of AI. Passing this bill in California, one of the world’s biggest AI hubs, could be a major step in that direction. Additionally, the bill would be the first actual United States AI safety regulation. 

Governor Newson has until September 30th to either sign, veto, or pocket sign the bill (pocket signing is a California policy where if the governor ignores the bill past its deadline, it becomes a law). 

In the meantime, he has signed five other AI bills into law just this past week. These include the following:

  • AB 2655 – requires large online platforms to remove or label election-related content that was digitally altered. This is the first law in the country to place this kind of responsibility onto social media platforms.
  • AB 2839 – prohibits anyone or any entity from knowingly distributing election-related deceptive content within 120 days of an election in California. This is the first law of this kind to actually go into effect (another has passed in the national Senate, but has since been stalled).
  • AB 2355 – requires that political ads that have been generated using AI to include a message that makes the viewer aware of the use of AI in the ad.

Additionally, AB2602 and AB1836 address concerns raised by the SAG during strikes last year.

  • 2602 – requires contracts to specify the use of AI-generated digital replicas of a performer’s voice or likeness, giving performers more control. The law also requires that performers be professionally represented while negotiating such contracts.
  • 1836 – protects the likeness of deceased performers, by prohibiting the commercial use of AI-generated replicas of their likeness without prior consent of their estates.

Leave a comment