California Transparency in Frontier Artificial Intelligence Act (SB 53)
Issued by
California Legislature; signed by Governor Gavin Newsom
- September 30, 2026 · Correction — Merged the duplicate SB 53 entry, which wrongly described the law as pending, and added the enacted thresholds, deadlines and penalties. (Cody Maxwell)
- October 1, 2026 · Correction — Rewrote the practical steps, newsletter hook, search description, and audience fields to match the entry's corrected content. (Cody Maxwell)
- October 2, 2026 · Correction — Corrected the 24-hour incident notice (to an appropriate authority such as law enforcement, not Cal OES) and limited the anonymous reporting channel duty to large frontier developers. (Cody Maxwell)
California's Transparency in Frontier Artificial Intelligence Act (SB 53) applies to developers of the largest AI models. Signed in September 2025 and in force since 1 January 2026, it requires published safety frameworks and transparency reports, and prompt reporting of critical safety incidents. The law also protects employees who report risks.
Applies To
Overview
Governor Gavin Newsom signed SB 53 on 29 September 2025, and it took effect on 1 January 2026. It applies to frontier developers, meaning developers that have trained a model using more than 10^26 operations of computing power. Large frontier developers, those with more than USD 500 million in annual revenue, carry the heaviest duties. Large frontier developers must publish a frontier AI framework. It explains how they assess and manage catastrophic risks, apply standards, and secure unreleased model weights. All frontier developers must publish a transparency report before deploying a new or substantially modified frontier model. Developers must report critical safety incidents to the California Office of Emergency Services within 15 days. Where an incident poses an imminent risk of death or serious injury, the developer must tell law enforcement or a public safety agency within 24 hours. The law also protects employees who report catastrophic risks, and it sets up CalCompute, a public computing cluster consortium. The Attorney General enforces the law, with civil penalties of up to USD 1 million per violation.
Key Requirements
- •Large frontier developers must publish and follow a frontier AI framework covering catastrophic risk assessment and mitigation.
- •Publish a transparency report before deploying a new or substantially modified frontier model.
- •Report critical safety incidents to the Office of Emergency Services within 15 days. If there is imminent risk of death or serious injury, tell an appropriate authority, such as a law enforcement or public safety agency, within 24 hours.
- •Do not retaliate against employees who report catastrophic risks. Large frontier developers must also provide an internal anonymous reporting channel.
- •Civil penalties of up to USD 1 million per violation, enforced by the Attorney General.
What Your Organization Must Do
- →Check whether any model you train exceeds 10^26 computing operations, which triggers frontier developer status.
- →Confirm whether annual revenue passes USD 500 million, which brings the heavier large frontier developer duties.
- →Publish and maintain a frontier AI framework covering catastrophic risk assessment, standards, and security of unreleased model weights.
- →Prepare a transparency report template and publish it before deploying each new or substantially modified frontier model.
- →Build an incident process meeting the 15 day Cal OES deadline, plus a 24 hour route to law enforcement or public safety agencies for imminent danger.
- →If you are a large frontier developer, set up an anonymous internal reporting channel, and brief managers that retaliation is prohibited.