The law library · California, United States · US state law
California SB 53, the Transparency in Frontier Artificial Intelligence Act
Signed by Governor Gavin Newsom on 29 September 2025 · key provisions from 1 January 2026
Signing date and effective date confirmed against multiple counsel analyses and the published bill record. Obligation detail is summarised rather than quoted, and we say so. The section on Executive Order N-9-26 was read at the signed order.
The first of the three American frontier-model laws, and the template the others followed. It does not reach your firm. It reaches the handful of companies that build the models your firm uses, and it makes them publish what they think could go catastrophically wrong.
What it requires
- RiskPublicly accessible documents assessing potential catastrophic risks from the developer’s AI models.
- StandardsPublished documentation of how the model incorporates national and international safety standards.
- WhoLarge frontier developers. Ordinary deployers are outside it.
- WhenSigned 29 September 2025; key provisions from 1 January 2026.
It is the first bill anywhere addressing catastrophic AI risk to become law, which is why it was reported as California assuming the role of lead US regulator of AI in the absence of a federal statute.
The three-state pattern
Read together, the American frontier-model laws are one instrument passed three times, each adding something:
- CaliforniaSB 53. Signed 29 Sep 2025, effective 1 Jan 2026. Published risk assessments and standards documentation.
- New YorkRAISE Act. Signed 19 Dec 2025, effective 1 Jan 2027. Published safety protocols.
- IllinoisSB 315. Signed 6 Jul 2026, effective 1 Jan 2027. Adds an independent third-party audit.
The trajectory is the point. California asks developers to publish what they believe. Illinois, nine months later, asks someone independent to check it. That is the same movement the courts have made, from accepting an attorney’s assurance to asking what they can show.
18 September 2026: independent verification, and a kill switch under study
On 18 September 2026 the Governor signed Executive Order N-9-26. It builds on SB 53 and on legislation signed earlier the same month, which the order describes as a framework for certifying independent verification organizations and regulating AI auditors; SB 813 on independent verification organizations was approved on 9 September 2026 as Chapter 179, Statutes of 2026, and AB 1405 on AI auditors the same day as Chapter 178. Government Code section 11549.82, named below, is the AB 1405 section that sets up the AI Auditor Registry. The order directs:
- by 1 May 2027, the Government Operations Agency to develop and post application requirements, procedures and criteria for independent verification organizations;
- by 1 December 2027, the same agency to complete the requirements of Government Code section 11549.82(a) and begin those of (b);
- by 16 November 2026, the agency, with the Governor’s Office of Emergency Services, to send the Governor recommendations on the feasibility and efficacy of amending state law to require that large frontier developers embed designated independent verification organizations onsite for periodic audits; that the safety frameworks, transparency reports and risk assessments frontier companies file “be independently verified”; the creation of a “kill switch” for frontier models, “with the efficacy of the switch verified on an ongoing basis by an independent verification organization”; and a definition of critical safety incidents that covers loss-of-control incidents.
What it does not do. The kill switch and onsite verifiers are subjects for recommendations due on 16 November 2026, not requirements. They concern large frontier developers only. The order states that it does not “create any rights or benefits, substantive or procedural, enforceable at law or in equity”. No California law today requires any company to have a kill switch.
The direction is the same one this page already describes: from what a developer publishes to what someone independent can check. Regulayer™ is not an auditor and no California law requires it. It is the control that holds an AI action until a named person’s authority is confirmed, and the sealed record an auditor can inspect afterwards.
Why it is on this list
Because a published claim invites the question of whether it held on a given day, and that question is answered with records. A framework document describes what a system is supposed to do. An audit, which Illinois now requires, asks what it actually did. Every organisation in scope of any of these three will eventually need evidence at the level of individual events, not policies.
What to do about it
This is an estate problem, not a document problem.
A published catastrophic-risk assessment is a claim about behaviour over time, and Illinois now requires an independent auditor to test it. An auditor asks for events, not policies. Regulayer™ records the decision as it is taken, outside the system being governed, content-free, and sealed.
Regulayer™ for enterpriseThe AI audit trail →
See also the law library and every jurisdiction we track.
Sources
- California SB 53, bill text, California Legislative Information
- Signed 29 September 2025; key provisions effective 1 January 2026. Confirmed across White & Case, Clifford Chance and Latham & Watkins analyses.
- Executive Order N-9-26, signed 18 September 2026, three pages, read in full
- SB 813, Independent verification organizations, bill history, California Legislative Information: approved by the Governor and chaptered 9 September 2026, Chapter 179
- AB 1405, Artificial intelligence: auditors: registration, chaptered text, California Legislative Information: approved by the Governor 9 September 2026, Chapter 178, read in full
- No section-by-section block: obligations here are summarised from counsel analysis rather than a section read of the enrolled bill, and we say so.
Related
- The law library
- New York · RAISE Act
- Illinois · SB 315
- California · AI Transparency Act
- California · AB 1405, the AI Auditor Registry
Executive Order N-9-26 read at source. Information, not legal advice.
Common questions
What is California SB 53?
The Transparency in Frontier Artificial Intelligence Act, signed on 29 September 2025. It makes California the first US state with a comprehensive legal framework for transparency, safety and accountability in the development and deployment of advanced AI models, and the first jurisdiction anywhere to legislate specifically about catastrophic risk from AI.
What does SB 53 require companies to do?
Large frontier developers must publish publicly accessible documents assessing potential catastrophic risks from their models, and publish documentation of how the model incorporates national and international safety standards.
When does SB 53 take effect?
It was signed on 29 September 2025, with key provisions taking effect on 1 January 2026.
How does SB 53 relate to New York's RAISE Act and Illinois SB 315?
They are the same pattern in three states. California legislated first, New York's RAISE Act was signed on 19 December 2025 and takes effect on 1 January 2027, and Illinois SB 315 was signed on 6 July 2026 and takes effect on 1 January 2027. Illinois adds a requirement none of the others has: an independent third-party audit.
Does SB 53 apply to a law firm or an ordinary business?
No. It reaches large frontier model developers, not businesses that deploy someone else's model. Its practical relevance to everyone else is that the developers you buy from now publish safety documentation you can ask to see.
Information, not legal advice, and not advice on the law of any jurisdiction. Every entry is verified against the issuing body’s own document; where a source is reporting rather than the document, we say so.