Some links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
The fight was real, but the headline’s “new law” was not: California’s SB 1047 was a 2024 bill, not an enacted law. The Legislature passed it, then Gov. Gavin Newsom vetoed it on September 29, 2024. The proposal would have imposed safety and security duties on developers of certain very large, powerful AI models—not blanket penalties every time a chatbot made a mistake. California later enacted a different frontier-AI law, SB 53, in 2025.
What was California’s SB 1047?
SB 1047, the Safe and Secure Innovation for Frontier Artificial Intelligence Models Act, was sponsored by state Sen. Scott Wiener during California’s 2023–24 legislative session. It targeted “frontier” models: highly capable systems whose development requires substantial computing resources. The bill’s scope was therefore narrower than all AI software or every chatbot interaction, though its criteria and obligations evolved during the legislative process. The official legislative record lists the bill as vetoed.
Contemporary summaries often described a threshold above $100 million in development cost, alongside computing-power criteria. That shorthand should not be treated as a complete statement of the final bill’s coverage. The enrolled text sets out the detailed definitions and requirements.
What the bill would have required
For covered models, SB 1047 proposed written safety and security protocols, safety testing and risk assessments, and a means to shut down or disable a model in an emergency. It also contemplated independent third-party audits in a future implementation period and added protections for whistleblowers reporting safety concerns. The “kill switch” label captured only one element of a broader safety framework.
Enforcement would have included civil actions by the California attorney general, including requests for injunctive relief and civil penalties for specified violations. The bill’s serious-harm framework was not a general ban on flawed outputs: it addressed harms such as death or bodily injury, property damage, theft or misappropriation, and imminent risks or threats to public safety. Penalties could be tied to the cost of computing power used to train a covered model. The precise legal conditions are in the bill text.
#1 Best Overall
The proposal chiefly concerned developers of covered frontier models and, in some circumstances, providers of computing power used to train them. That is different from making a company automatically liable for every output generated by a customer. In AI accountability debates, responsibility can fall on the user who misuses a system, the developer who designs it, the deployer that puts it into a consequential setting, or more than one party depending on who controlled the relevant risk. SB 1047 was notable for seeking duties beyond user-only responsibility for the most powerful models.
Why AI companies objected
OpenAI and other opponents argued that the proposal could burden innovation, discourage investment, and prompt engineers or startups to leave California. OpenAI’s position, reported at the time by Futurism, framed the bill as a threat to the state’s AI economy. Those were warnings about possible consequences, not proof that companies would have relocated.
Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallOutdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchCritics also said that obligations designed for large AI laboratories could create uncertainty for open-source developers, smaller organizations, and people who modify or redistribute models. An open model may be changed or deployed by parties far removed from its original developer, complicating questions about who could anticipate or prevent misuse. Opponents further warned that a California-specific regime could fragment rules across states and preferred a more uniform federal framework.
Liability was another point of contention. A developer might build safeguards, yet a malicious user could bypass them or a downstream business could integrate the model into a risky product. Critics worried about exposure for unpredictable downstream harms. That concern deserves context: the bill did not simply impose automatic liability for any hallucination or offensive answer. It created specified duties and enforcement provisions focused on serious harms and public-safety risks.
Rank #3
What supporters wanted
Supporters argued that the most powerful models could help users carry out serious cyberattacks, biological threats, fraud, or other large-scale harms. Developers have access to training information and testing resources that ordinary users do not, they said, and should be expected to assess foreseeable risks and maintain safeguards. They also argued that voluntary commitments might not be enough when commercial pressure rewards rapid deployment.
The dispute was not simply “companies versus safety.” It was about which systems merit regulation, whether model scale is a sound proxy for risk, how much responsibility belongs to developers versus deployers and users, and whether state rules would help or hinder innovation. A general-purpose model might be relatively safe in one setting but dangerous when integrated into health care, finance, infrastructure, or cybersecurity. Conversely, a model can generate harmful content without causing the severe harm contemplated by the proposal. Those distinctions make both risk assessment and legal responsibility difficult.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problemsDid SB 1047 punish companies whenever AI did something bad?
No. A wrong, biased, or rude chatbot response is not by itself the same as a statutory serious harm or imminent public-safety threat. The proposal focused on covered models, required safety practices, and civil enforcement tied to specified conditions. It was not a blanket rule that every bad output would trigger a fine.
Nor would a shutdown mechanism solve every safety problem: disabling a hosted service cannot necessarily retract model copies already downloaded or deployed elsewhere. Testing can also miss misuse scenarios, and compliance documentation does not guarantee that safeguards work in practice. These are practical limits of any model-centered approach, not reasons to equate the bill with ordinary content moderation.
Why Newsom vetoed it
Newsom vetoed SB 1047 on September 29, 2024. In his veto message, he called the bill well-intentioned but said it was not the best approach. His central objection was that it focused on the size of a model rather than adequately distinguishing AI used in high-risk settings, critical decision-making, or with sensitive data. The argument reflects a different regulatory emphasis: focus on how and where AI is used, not only on the resources needed to build it.
That does not mean Newsom rejected AI safeguards as a goal. His administration announced other safe-and-responsible-AI initiatives after the veto, as described in the governor’s announcement.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
What happened next: California’s SB 53
The policy debate continued. On September 29, 2025, Newsom signed SB 53, the Transparency in Frontier Artificial Intelligence Act. The governor described it as a framework for transparency and online safety around frontier AI while supporting continued innovation. SB 53 is a later law with a different title and framework; it is not SB 1047 under another name. See the signing announcement.
So the practical answer is: SB 1047 passed California’s Legislature but never took effect because it was vetoed. The headline captured a genuine 2024 clash, but called a bill a law and compressed a dispute over frontier-model safety, liability, and regulatory design into “bad stuff.” California later pursued a different legislative approach with SB 53.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
