Law / United States / 
California
Transparency in Frontier Artificial Intelligence Act (SB 53)
Cal. Bus. and Prof. Code Sections 22757.10 to 22757.16
In force since .
An AI governance rule binding private bodies.
- Obligation class
- Disclosure, Governance, Reporting
- Audit expectation
- periodic
- Who audits it
- Regulator, Self
- Where the report goes
- Filed with regulator, Published summary
What it requires
- If you are a large frontier developer (a frontier developer whose group had annual gross revenues over $500,000,000 in the prior calendar year), write, implement and publish on your website a frontier AI framework describing how you define and assess catastrophic-risk thresholds for your frontier models and apply mitigations, and review that framework at least once a year
- If you are a large frontier developer, make your published frontier AI framework describe your cybersecurity practices to secure unreleased model weights from unauthorized modification or transfer by internal or external parties (Section 22757.12(a)(7))
- Before or when you deploy a new or substantially modified frontier model, publish a transparency report giving the model's release date, supported languages, output modalities, intended uses, and any generally applicable restrictions
- If you are a large frontier developer, add to that transparency report summaries of your catastrophic-risk assessments, their results, and the extent of any third-party evaluator involvement
- If you are a large frontier developer, transmit summaries of catastrophic-risk assessments from your frontier models' internal use to the California Office of Emergency Services on a quarterly or agreed schedule
- If you are a frontier developer, report a critical safety incident pertaining to one or more of your frontier models to the Office of Emergency Services within 15 days of discovering it; if you discover information about the incident after filing, you may file an amended report (Section 22757.13(c))
- If you are a frontier developer, disclose a critical safety incident that poses an imminent risk of death or serious physical injury within 24 hours of discovering that it does so, to an authority, including any law enforcement agency or public safety agency with jurisdiction, that is appropriate based on the nature of the incident and as required by law (Section 22757.13(c)(2))
- Do not make a materially false or misleading statement about catastrophic risk from your frontier models, or, if you are a large frontier developer, about your compliance with your own frontier AI framework
- Do not retaliate against a covered employee who discloses, in good faith and with reasonable cause, that your activities pose a catastrophic-risk danger to public health or safety or that you violated this Act; if you are a large frontier developer, also provide an internal channel for anonymous disclosure of that kind
These lines are LexLint's reading of the law's text.
In the frameworks
- NIST AI RMF
- GOVERN 1.4The risk management process and its outcomes are established through transparent policies, procedures, and other controls based on organizational risk priorities.
- GOVERN 1.5Ongoing monitoring and periodic review of the risk management process and its outcomes are planned and organizational roles and responsibilities clearly defined, including determining the frequency of periodic review.
- NIST AI 600-1
- GAI-RISK-07Human-AI Configuration
- NIST AI 200-2 Draft
- OWASP LLM Top 10
- LLM03:2026Excessive Agency
- OWASP Agentic Top 10
- ASI09Human-Agent Trust Exploitation
- MIT mitigations Draft
- FINOS AIGF
- AIR-PREV-005System Acceptance Testing
- AIR-PREV-006Data Quality & Classification/Sensitivity
- AIR-PREV-024Human-in-the-Loop Action Approval Gate
- AIR-DET-004AI System Observability
- AIR-DET-011Human Feedback Loop for AI Systems
- AIR-DET-021Agent Decision Audit and Explainability
- ETSI EN 304 223
- 5.1.4-1Capability for human oversight designed in and maintained.
- 5.2.1-1Full inventory of assets, including their dependencies and connections.
- 5.2.4-1Developers keep an audit trail of system design and post-deployment maintenance plans and offer it downstream.
- 5.4.2-1Operators log system and user actions.
Our reading of where this law's duties sit in each framework. It never says that running a control meets the law. How we read a framework against the law
Identifiers refer to ETSI EN 304 223 V2.1.1 (2025-12), © ETSI 2025, all rights reserved. ETSI owns the standard; no ETSI text (titles or sentences) is reproduced here. The standard is free to download from ETSI.
If you get it wrong
Criminal exposureNo
Private right of actionYes
Penalty structure
Bus. and Prof. Code section 22757.15(a): a large frontier developer that fails to publish or transmit a required document, makes a false or misleading statement in violation of section 22757.12(e), fails to report a critical safety incident, or fails to comply with its own frontier AI framework is subject to a civil penalty, scaled to the severity of the violation, of up to $1,000,000 per violation. Section 22757.15(b) limits recovery to a civil action brought by the Attorney General alone.
- Rule
- Per violation only
- As of
- Currency
- USD
- Per violation unit
- Violation
- Per violation amount
- 1,000,000
Who enforces it
Enforcement body
The chapter's own civil penalty (Section 22757.15) is recovered only in a civil action brought by the Attorney General. The whistleblower-retaliation protections in Labor Code Section 1107 et seq. are enforced separately, through a covered employee's own civil action.
Enforcement record
The Attorney General's own dedicated page for the Transparency in Frontier Artificial Intelligence Act describes the covered-employee whistleblower reporting channel and the annual aggregated whistleblower-report disclosure required by Section 22757.14(d), a report not yet due, but names no civil action or civil penalty brought under the chapter's own enforcement provision, Section 22757.15. A review of the Attorney General's most recent press releases, September 1 through , within the chapter's to date effective period, names no frontier developer, catastrophic-risk finding, or Section 22757.15 penalty; earlier weeks of the chapter's effective period were not reviewed.
- As of
- Source link
- https://oag.ca.gov/sb53
Settledness
- As of
- Open questions
- Does the quantity of computing power "greater than 10^26 integer or floating-point operations" in section 22757.11(i) count the original training run of a model that a different developer later fine-tunes?
- Does "frontier developer" in section 22757.11(h) reach a person who "intends to use" at least as much computing power as the threshold but has not yet trained a frontier model?
What it makes you log
Who may demand the log
Regulator, Public
What the log must hold
Event time, System identity, Decision basis
Log retention
Section 22757.12(f)(2)'s five-year floor applies to the unredacted version of information a frontier developer redacts from a published framework or transparency report; it does not apply to a document the developer never redacts.
- Unit
- Years
- As of
- Basis
- Fixed
- Minimum value
- 5
Logging duty
A large frontier developer must publish a frontier AI framework and a transparency report, and every frontier developer must report a critical safety incident to the Office of Emergency Services stating its date, the reasons it qualifies, a description, and whether it involved internal use. Section 22757.13 never uses the words log, record, event recording, or audit trail; it uses report, and section 22757.12(f) uses document. Filing an accurate incident report, and the transparency report's own summaries of risk assessments, cannot be produced on demand without an underlying record the developer already keeps.
- Kind
- Implicit
- As of
- Provision
- Sections 22757.12 and 22757.13
- Trigger
- security_incident
Why the legislator wanted it
SEC. 1(i) of the enacted bill, part of the Legislature's own findings and declarations that precede the chapter's operative text.
- As of
- Quote
- Incident reporting systems enable monitoring of the post-deployment impacts of artificial intelligence.
- Source link
- https://leginfo.legislature.ca.gov/faces/billTextClient.xhtml?bill_id=202520260SB53
- Source kind
- Recital
- Purpose
- incident_reconstruction
- oversight_and_correction
What this law does
California's Transparency in Frontier Artificial Intelligence Act (TFAIA) was added by SB 53, chaptered as Stats 2025 Chapter 138, and is effective per the code's own history note.
A large frontier developer (a frontier developer whose group had annual gross revenues over $500,000,000 in the preceding year) to write, implement and publish on its website a frontier AI framework describing how it defines and assesses catastrophic-risk thresholds for its frontier models (foundation models trained using more than 10^26 integer or floating-point operations) and applies mitigations, and to review that framework at least annually.
The framework must also describe the developer's cybersecurity practices to secure unreleased model weights from unauthorized modification or transfer by internal or external parties. Before or when deploying a new or substantially modified frontier model, every frontier developer must publish a transparency report with baseline model information, and a large frontier developer must add summaries of its catastrophic-risk assessments and any third-party evaluator involvement.
A large frontier developer must also transmit summaries of internal-use catastrophic-risk assessments to the Office of Emergency Services on a quarterly or agreed schedule, and every frontier developer must report a critical safety incident to the Office within 15 days of discovering it, or 24 hours if it poses an imminent risk of death or serious injury.
A frontier developer may not make a materially false or misleading statement about catastrophic risk or, for a large frontier developer, about its framework compliance, and a companion whistleblower chapter (Lab.
Code Section 1107 et seq.) bars a frontier developer from retaliating against a covered employee who reports a reasonable belief of catastrophic-risk danger or a TFAIA violation, arming that employee with a private civil action and attorney's fees, separate from the chapter's own civil penalty which the Attorney General alone may bring.
When LexLint raises it
When your app profile says your app generates content with AI or is a frontier developer.
Back to the example · Lint your app
Page updated .