California AI Rules: What Newsom’s Plan Would Change

Last updated November 19, 2026
Governor Gavin Newsom proposed tougher rules for powerful artificial intelligence on November 18, 2026. Here is what would change and what comes next.
- The plan is a set of recommendations for state lawmakers, not a law yet.
- It targets the biggest developers of advanced artificial intelligence models, not everyday users.
- Ideas include outside safety checks, 48-hour incident reports and emergency shutdown power.
- Supporters call it a needed safeguard. Critics question its politics and its effect on innovation.
What Newsom Proposed
The state Government Operations Agency released an 85-page report on artificial intelligence safety for 2026 and beyond. You can read the full report on the agency website.
Newsom presented it in Washington and pitched it as a model for national rules, according to his official announcement.
The report builds on Senate Bill 53, a 2025 California law on frontier model safety. It also follows an executive order Newsom signed in September 2026.
What Would Change for Big Artificial Intelligence Companies
| Area | Today under current law | Under the proposal |
|---|---|---|
| Safety standards | Companies publish their own safety plans | State sets binding baseline standards |
| Outside checks | Not required on an ongoing basis | Licensed independent reviewers check developers |
| Incident reports | Within 15 days for most incidents | Within 48 hours, or 24 hours if danger is imminent |
| Emergency action | No state shutdown power | State can order a dangerous model shut down |
| Public computing | Planned state computing project | Funded partly by a fee on reviews |
Details come from the state safety report dated November 16, 2026.
Developers could challenge most state orders in court before they take effect. Emergency shutdown orders would stay in force while a court reviews them, the report says.
What It Means for Everyday Users
The rules would apply to a small number of very large developers. Most people would not see a direct change in the apps they use.
Supporters say the main benefit is fewer serious failures, such as models helping with cyberattacks. Critics say heavy rules could slow new features or push work out of state.
Should states set their own rules for advanced artificial intelligence?
- Yes, states should act now
- No, it should be national rules only
- Only with light rules
- Not sure
Who Supports the Plan
Anthropic, a major developer, said it was encouraged by the call for independent evaluations, the Sacramento Bee reported.
Governor-elect Xavier Becerra called the report a starting point and said he would work with lawmakers on it. The California Labor Federation also backed strong rules.
Free updates
Join the Digieh list
Be the first to know when our weekly brief launches.
No spam. Unsubscribe anytime. Read our Privacy Policy.
What Critics and Skeptics Say
A Daily Caller opinion piece argued the move is mainly about Newsom’s national ambitions ahead of 2028.
Industry groups have pushed back on similar ideas before. Newsom vetoed a 2024 safety bill. Its critics had warned it would stifle innovation, CBS News reported.
Some officials in Washington also prefer national rules over a patchwork of state laws. That debate could shape what California passes.
| Argument | Supporters say | Critics say |
|---|---|---|
| Safety | Outside checks catch risks companies miss | Companies already test models carefully |
| Innovation | Clear rules build public trust | Strict rules could slow progress |
| Who should regulate | States must act while Congress waits | National rules avoid a patchwork |
| Politics | The plan responds to real incidents | The timing serves a 2028 campaign |
Arguments are a Digieh summary based on the Sacramento Bee, Daily Caller and CBS News reports linked above.
What Happens Next
Lawmakers return in December 2026 and could start drafting bills. Any final law would likely be signed by Becerra, who takes office in January.
Curious how these tools work today? Our explainer on agents and chatbots covers the basics. You can also read about Gemini 4 Neon, which Google says includes strict safety tests.
Security problems on the web are a related worry. Our news on Nginx security flaws shows how quickly fixes must move.
Test Yourself
What is the proposed deadline for most critical incident reports?
- 15 days
- 48 hours
- 30 days
The report recommends cutting most reporting deadlines from 15 days to 48 hours.
Frequently Asked Questions
Is Newsom’s artificial intelligence plan a law?
No. It is a set of recommendations. Lawmakers must pass bills and the governor must sign them.
Who would the rules apply to?
They would apply to large developers of the most advanced models, not to small businesses or everyday users.
What is an artificial intelligence kill switch?
It refers to state power to order an emergency shutdown of a model that poses an imminent catastrophic risk.
Editor’s Note
This article is accurate as of November 19, 2026. Details come from the state report, the governor’s office, the Sacramento Bee, the Daily Caller and CBS News. Digieh will update this story as bills move forward.
Spot an error in this article? Please contact us at support@digieh.com or read our Corrections Policy.
Keep exploring
Your next read is waiting
1Education & CareersHow to Take Notes With the Cornell Method8 min read
2Parenting & FamilyHealthy Lunchbox Ideas Your Kids Will Actually Eat7 min read
3InsuranceHome Insurance Claim Steps That Protect Your Payout8 min read
4RecipesEasy Chicken Karahi Recipe Made in One Pan7 min read
5Fitness & ExerciseDesk Stretches to Ease Neck and Back Stiffness7 min read
6TechnologyHow to Make a GIF From Video, Fast5 min readFound this useful?












