What California SB 53 requires from frontier model developers
California Transparency in Frontier AI Act (SB 53)
- Status
- In force
- Binding
- Yes
- Object analysed
- AI model
- Requirements
- 5
- Next milestone
- Jan 1, 2027
In short
Transparency and safety for frontier models: published safety framework, transparency reports, critical incident reporting, whistleblower protection.
Steps to compliance
- Qualify each AI systemAxes to decide: Developer status.
- Determine your roleDuties vary by role: Frontier developer.
- Apply the 5 requirementsThey focus on: Governance & accountability, Risk management, Cybersecurity and Transparency & notice.
- Prove it with checks7 checks to document, 4 of which also serve AI Act, PL 2338 and AI Basic Act.
- Track the deadlinesNext milestone: Jan 1, 2027, First anonymised annual OES reports (set in the text).
Scope and penalties
- Kind
- Statute
- Scope
- Frontier developers (> 10^26 FLOP); stronger duties above $500M revenue.
- Territorial reach
- Models made available in California.
- Penalties
- Up to $1M per violation; Attorney General enforcement.
- Jurisdiction
- California
Timeline
Qualifying a system
Classification axes and possible verdicts
Developer status
Requirements
5 requirements
| Code | Article | Requirement | Applies to | Checks |
|---|---|---|---|---|
| SB53-01 | §22757.12(a) | Publish a frontier AI framework (thresholds, mitigations, governance, weight security) | ||
| SB53-02 | §22757.12(c) | Transparency report before each new model | ||
| SB53-03 | §22757.13 | Report critical safety incidents to OES within 15 days (24h if imminent) | ||
| SB53-04 | §22757.12(d) | Quarterly summary of catastrophic risk assessments | ||
| SB53-05 | Lab. Code §1107 | Whistleblowers: protection and anonymous internal channel |
Checks to document
Evidence collected for a check counts for every regulation that uses it.
| Code | Check | Scope | Themes | Also used by |
|---|---|---|---|---|
| VER-032-G-02 | Systemic risks assessed and mitigation measures documented | Model | ||
| VER-032-G-04 | Cybersecurity of the model and physical infrastructure ensured | Model | ||
| NEW-US-CA-SB53-01 | Frontier AI framework published and reviewed yearly proposed | System | ||
| VER-031-G-02 | Documentation for downstream providers compliant with Annex XII | Model | ||
| VER-032-G-03 | Serious incidents documented and reported to the AI Office | Model | ||
| NEW-US-CA-SB53-02 | Quarterly summary sent to OES proposed | System | ||
| NEW-US-CA-SB53-03 | Anonymous internal reporting channel in place proposed | System |
Themes covered
Frequently asked questions
Who is in scope of SB 53?
Frontier developers (> 10^26 FLOP); stronger duties above $500M revenue. Models made available in California.
What penalties does SB 53 carry?
Up to $1M per violation; Attorney General enforcement.
When do the SB 53 obligations apply?
Sep 29, 2025: Signed into law; Jan 1, 2026: Effective date; Jan 1, 2027: First anonymised annual OES reports.
Is SB 53 binding?
Yes. Kind: statute. Status: in force.
How does SB 53 relate to other regulations?
The same checks serve several texts. Shared checks: AI Act (4), PL 2338 (2) and AI Basic Act (2).
Related regulations
Official sources
Data checked on Sep 25, 2026. General information, not legal advice. Check the official texts and get advice for your situation.
Run these requirements across all your AI systems
TrustFlow inventories your systems, qualifies them under each regulation and collects evidence once for every referential.