The Clock Runs Out on 30 AI Bills: California's Deadline Day Scorecard

Wednesday, September 30, 2026 is decision day for artificial intelligence policy in the United States’ largest state. Under Article IV, Section 10 of the California Constitution, Governor Gavin Newsom has until midnight to sign or veto the roughly 30 AI-related bills the legislature shipped to his desk when the session ended on August 31 — and any bill he neither signs nor vetoes becomes law automatically, without his signature.

It is the final act of the most consequential AI legislative season any U.S. state has ever had. And it closes with a paradox: a governor who has signed a historically aggressive package of AI safeguards while simultaneously rejecting — for the third year running — the one measure that would have imposed hard pre-deployment controls on frontier model developers.

What has already been signed

The audit regime (September 9). Newsom signed SB 813 (McNerney), creating the nation’s first framework for “independent verification organizations” (IVOs) — accredited third parties that can assess the safety of frontier AI systems. Companion bill AB 1405 established an AI auditor registry. Together they give California something no other state has: an end-to-end enforcement architecture built on disclosure duties, accredited auditors, and mandated audits.

The child-safety package (September 10). Thirteen bills, anchored by SB 1119 — Adam’s Law, the most comprehensive companion-chatbot safety statute in the country. The package bans addictive feeds and autoplay for users under 16, criminalizes AI-generated child sexual abuse material, requires independent child-safety audits of chatbots, and creates civil liability of up to $1 million per child for negligent large social platforms.

The kill-switch executive order (September 18). Executive Order N-9-26 directs state agencies to accelerate the IVO regime and to formally explore an emergency “kill switch” for frontier AI models — a mechanism capable of halting a system that resists shutdown. Recommendations on whether state law should mandate one are due November 16, 2026, and the auditor certification rules must land by May 1, 2027. On September 23, Newsom named a panel of outside experts charged with delivering on the order. Senator Scott Wiener, whose 2024 SB 1047 first proposed kill-switch requirements (and was vetoed), endorsed the move as reviving the core of his bill through executive action.

Privacy expansions (September 27–28). SB 923 (Becker), signed Sunday, expands California’s data-deletion rights — a direct counterweight to AI companies’ voracious appetite for training data.

What got vetoed

The highest-profile casualty is AB 1542, which would have amended the California Consumer Privacy Act to prohibit the sale or sharing of sensitive personal information — precise location, health, and immigration-related data. Newsom vetoed it on September 28, with Consumer Reports noting the decision leaves California’s privacy floor behind five other states. It is a reminder that even in a maximalist session, the governor’s line remains roughly: audits and transparency, yes; outright prohibitions on data commerce, no.

And hovering over everything is the 2024 veto of SB 1047 — Wiener’s frontier-AI bill that would have required pre-release safety testing, third-party audits, and a kill switch for the largest models. Mission Local reported earlier this month that the narrower law that passed instead, SB 53 (2025), did not cover the recent OpenAI hack — a gap Wiener himself calls a mistake, and part of why the kill-switch idea is now back on the table via executive order rather than statute.

The bills going down to the wire

As of September 28, several major AI bills remained unsigned — meaning they either get a late signature or quietly become law without one:

• SB 947 (McNerney) — the “No Robo Bosses Act.” From July 1, 2027, employers could not rely solely on an automated decision system to discipline or terminate a worker; a human would have to corroborate the outcome. Newsom vetoed its predecessor, SB 7, in 2025 as overbroad, and business groups mounted an organized veto campaign this year.
• AB 1883 (Bryan) — workplace AI surveillance. Would prohibit AI-powered workplace surveillance tools that track employees’ emotional states or biometric-derived metrics, with safety-related exceptions.
• SB 951 (Reyes) — AI layoff notice. Extends the Cal/WARN Act so that mass layoffs or relocations caused by technological displacement trigger advance notice obligations to affected workers.
• SB 760 and SB 1083 — further kids’ chatbot safety and consumer protection measures that had not yet received action.

Why this session mattered

Three structural facts make the 2026 session a turning point. First, volume: roughly 30 AI bills reached the governor’s desk from a single two-year session, up from a handful two years ago. Second, architecture: with SB 53 (frontier transparency, 2025), SB 813/AB 1405 (auditors, 2026), and Adam’s Law (chatbot safety, 2026), California now has an interlocking system covering disclosure, verification, and litigation — the closest thing to the EU AI Act’s enforcement model on U.S. soil. Third, the federal vacuum: Newsom explicitly framed the September 9 signings as a call for Washington “to do its part,” and the kill-switch executive order is designed to survive federal inaction.

The pattern in Newsom’s decisions is consistent and worth watching for anyone building in AI: he signs measures that create oversight infrastructure (auditors, audits, transparency, notice) and vetoes or stalls measures that mandate specific product engineering (pre-deployment testing requirements, kill switches in statute, data sales bans). The kill-switch question now moves to the November 16 expert recommendations — and if the panel endorses a statutory mandate, 2027 will reopen the fight Wiener lost in 2024.

By midnight tonight, every one of the 30 bills will have a final answer. For the AI industry, California’s compliance floor just became the national one.