Every AI Policy Ever · open in the interactive register

LapsedProposed · Bill / draft law

Safe and Secure Innovation for Frontier Artificial Intelligence Models Act (SB 1047)

Record US-CA-03 · California Senate Bill 1047 (2023-2024 Regular Session)

Issued by
California State Legislature (vetoed by Governor Gavin Newsom)
Jurisdiction
California (US)
Level · region
Sub-national · North America
Country
United States
Published / adopted
2024-09-29
Language
English
Official identifier
California Senate Bill 1047 (2023-2024 Regular Session)

Summary

Bill passed by the Legislature in August 2024 that would have required developers of the most advanced AI models, those costing over $100 million to train, to guard against critical harms: before training, protect models against unauthorised access and misuse, build in the capability for a full shutdown and adopt a written safety and security protocol, with third-party audits and AI safety incidents reported within 72 hours. It would also have created the Board of Frontier Models, whistleblower protections for employees of frontier AI laboratories and the CalCompute public computing cluster. The Governor vetoed it.

Safety topics

general-purpose / frontier AIcatastrophic riskevaluation & testingincident reportingaccountability & liability

Status notes

Sent to the Governor after the Senate's 29-9 vote announced on 29 Aug 2024, and vetoed on 29 Sep 2024 (date_published is the veto date). The veto message objects that the bill covered only the most expensive and large-scale models, whether or not a system is deployed in high-risk environments, involves critical decision-making or uses sensitive data, warns that smaller, specialised models may prove as dangerous, and asks for an approach informed by an empirical trajectory analysis of AI systems and capabilities. The same day the Governor announced that Fei-Fei Li, Mariano-Florentino Cuéllar and Jennifer Tour Chayes would help lead California's effort to develop guardrails for generative AI; their report followed in June 2025 (US-CA-04). Included for historical context.

Sources

Open the official source https://www.gov.ca.gov/wp-content/uploads/2024/09/SB-1047-Veto-Message.pdf

Local copy

Archived. Last attempt 2026-10-17 23:05:41 UTC

Official source (third-party copy, not published) archive/US-CA-03/2026-10-11/official.pdf
Retrieved 2026-10-11 14:04:44 UTC
SHA-256 fe604e68446fc6db1ea1f4af6078d74d3feb184256147a4c83bcdbdfc77f83c2

Verification

A
The instrument's own official text, or an official register, was stored and confirms the title, date and status. checked 2026-10-18 · official-url-fetched
Recheck of 18 Oct 2026. Stored and read: the Governor's veto message (gov.ca.gov PDF, stamped 29 Sep 2024), which gives the bill number, the veto, the Board of Frontier Models and the reasons; the Governor's release of 29 Sep 2024 (gov.ca.gov), which announces the veto and the experts; Senator Wiener's release of 29 Aug 2024 (sd11.senate.ca.gov), which gives the final Senate vote, the $100 million threshold, CalCompute and the whistleblower protections; and the Assembly Privacy and Consumer Protection Committee's analysis (hearing of 18 Jun 2024, apcp.assembly.ca.gov), which gives the title as its subject and describes the full-shutdown, safety and security protocol, third-party audit, 72-hour incident-reporting and Board duties. That analysis describes the bill as amended on 5 Jun 2024 and proposed to be amended, not the enrolled bill the Governor vetoed, so the summary's specific duties rest on an earlier version (see open questions). The enrolled text on leginfo.legislature.ca.gov refused automated downloads (HTTP 403).
Second check: Independent check by Claude Code (claude-sonnet-5-5), 2026-10-18: 7 field(s) supported.

History

  1. 2026-10-18notes: Passed by the Legislature in Aug 2024 and vetoed on 29 Sep 2024 (date_published → Sent to the Governor after the Senate's 29-9 vote announced on 29 Aug 2024, and — the August passage now rests on the author's release (final Senate vote 29-9, 29 Aug 2024); the Governor's reasons rewritten from the veto message; 'shaped SB 53' replaced by the announcement of the experts the same day and a link to their report (US-CA-04)2026-10-18-recheck-north-america-2 · Claude Code (claude-opus-5-5)
  2. 2026-10-18summary: Passed bill that would have required developers of very large (covered) AI model → Bill passed by the Legislature in August 2024 that would have required developer — rewritten from the stored sources: the $100 million threshold, the duties before training, third-party audits, 72-hour incident reporting, whistleblower protections and CalCompute added; the veto stated2026-10-18-recheck-north-america-2 · Claude Code (claude-opus-5-5)

Related records

California SB 53 (TFAIA) California (US) · In force

California Frontier AI Policy Report California (US) · Published (non-binding, current)

Citation

California State Legislature (vetoed by Governor Gavin Newsom). Safe and Secure Innovation for Frontier Artificial Intelligence Models Act (SB 1047) [California Senate Bill 1047 (2023-2024 Regular Session)]. 2024-09-29. https://www.gov.ca.gov/wp-content/uploads/2024/09/SB-1047-Veto-Message.pdf. Status: Lapsed; verified 2026-10-18 (level A). Local copy archive/US-CA-03/2026-10-11/official.pdf retrieved 2026-10-11 14:04:44 UTC, SHA-256 fe604e68446fc6db1ea1f4af6078d74d3feb184256147a4c83bcdbdfc77f83c2. Every AI Policy Ever, version 1.18 (snapshot 2026-10-18), record US-CA-03.

Improve this record

Know of an amendment, a status change or an error? Report news or an error on GitHub; the form opens with this record filled in.