Skip to main content
NOPE
Failed Law AI Safety

CA SB 1047

California SB 1047 (Safe and Secure Innovation for Frontier AI)

Would have required safety testing and "kill switch" capabilities for frontier AI models above certain compute thresholds. Governor vetoed citing concerns about threshold-based approach.

Jurisdiction

California

Enacted

Pending

Effective

TBD

Enforcement

TBD

Vetoed September 2024

CA Legislature

Why It Matters

Clearest US example of attempted frontier model safety regulation. Veto reasoning (prefer risk-based over size-based) may shape future attempts.

Recent Developments

Governor Newsom's veto message cited concerns that size-based thresholds were not risk-informed and could create false security.

At a Glance

Applies to

Foundation Model

Who Must Comply

  • Would have applied to frontier AI models exceeding compute thresholds

Obligations fall on:

Safety Provisions

  • Would have required: safety testing before deployment
  • Would have required: ability to shut down models
  • Would have required: third-party audits for largest models

View on map

California

Focus Areas

General regulation

Cite This

APA

California. (n.d.). California SB 1047 (Safe and Secure Innovation for Frontier AI).

Related Regulations

Enacted US-CA

CA AB 1405

Requires AI auditors to register with the Government Operations Agency and bars unregistered persons from offering, selling or conducting a covered AI audit from January 1, 2029. Registered auditors must maintain independence from the auditee, follow applicable audit protocols, report in detail, and keep records for at least 10 years. Certified public accountants licensed by the California Board of Accountancy are treated as satisfying the independence requirements if they comply with applicable professional standards.

Enacted US-CA

CA SB 813

Creates a voluntary state designation for independent verification organizations: AI auditors that the Government Operations Agency recognises as having demonstrated expertise in assessing the risks posed by an AI system or model. The agency must establish application requirements, designation criteria and suspension procedures on or before January 1, 2028, and designated organizations report annually to the Legislature. Engaging a designated organization is not mandatory for AI developers or deployers.

In Effect US-IL

IL AI Employment Law

Amends Illinois Human Rights Act to make it a civil rights violation for employers to use AI that discriminates based on protected classes in employment decisions. Requires notice to employees and applicants when AI is used in hiring, firing, promotion, or discipline.

In Effect US-HI

HI Deepfake Civil Remedy Act

Creates a civil cause of action allowing individuals to sue over realistic digital imitations (deepfakes) depicting them without consent, with statutory damages available.

In Effect US-ND

ND Healthcare AI Act

Regulates AI use in healthcare prior authorization decisions. Requires adverse determinations be reviewed by licensed physicians qualified in relevant specialty. Sets decision timelines: 7 days for non-urgent, 72 hours for urgent services. Prohibits prior authorization for emergency services and opioid use disorder treatment.

In Effect US-TX

TX Social Media Deepfake Takedown

Requires social media platforms to provide accessible complaint system for explicit deepfake material. Platform must confirm receipt within 48 hours, investigate within 30 days (60 days if delayed), and provide mandatory updates to reporting user.

All regulations See what's likely relevant to your deployment

Last updated January 22, 2026. Verify against primary sources before relying on this information.