| CA
SB 53: Transparency in Frontier Artificial Intelligence Act
California SB 53 is the baseline every other state is measured against. |
- Frontier developer: trains a model above 10^26 integer or floating-point operations.
- Large frontier developer: that, plus annual gross revenues above $500,000,000 with affiliates.
|
- Catastrophic risk: more than 50 deaths or serious injuries, or more than $1,000,000,000 in property damage, from a single incident of:
- expert-level assistance in creating or releasing a chemical, biological, radiological or nuclear weapon;
- conduct with no meaningful human oversight that is a cyberattack, or would be the crime of murder, assault, extortion or theft;
- a model evading the control of its developer or user.
|
- Publish a frontier AI framework. First.
- Transparency report at every deployment.
- Critical safety incident reported to the Office of Emergency Services in 15 days.
- 24 hours where risk of death or serious injury is imminent.
- Quarterly internal-use catastrophic-risk summaries.
- Whistleblower protection.
|
In force since 1 January 2026. |
| NY
RAISE Act, as amended by S8828: Gen. Bus. Law art. 44-B
An SB 53 copy, plus a registration regime run by the state's financial regulator, plus a 72-hour incident clock, minus whistleblower protections. |
- Frontier model: trained above 10^26 integer or floating-point operations.
- Large frontier developer: a frontier developer that, with its affiliates, had annual gross revenues above $500,000,000 in the preceding calendar year.
|
- Catastrophic risk: more than fifty deaths or serious injuries, or more than one billion dollars ($1,000,000,000) in property damage, from a single incident of:
- expert-level assistance in creating or releasing a chemical, biological, radiological or nuclear weapon;
- conduct with no meaningful human oversight that is a cyberattack, or would be the crime of murder, assault, extortion or theft;
- a model evading the control of its developer or user.
|
- Publish a frontier AI framework.
- Transparency report at deployment.
- Critical safety incident reported to the office in 72 hours.
- Quarterly internal-use catastrophic-risk summaries.
- Register a disclosure statement with the Department of Financial Services before developing or deploying.
- Pay assessments that fund the regulator.
|
Enacted; effective 1 January 2027. |
| IL
Public Act 104-0538 (SB 315): Artificial Intelligence Safety Measures Act
An SB 53 copy, plus first-in-the-nation annual independent audits, plus a 72-hour incident clock. |
- Frontier developer: trains a model above 10^26 integer or floating-point operations.
- Large frontier developer: that, plus annual gross revenues above $500,000,000 with affiliates, in the preceding calendar year.
|
- Catastrophic risk: 50+ deaths or serious injuries, or $1B+ in property damage, in a single incident.
- A reportable critical safety incident is any of:
- weight theft causing death or injury;
- realized catastrophic risk;
- loss of control causing death or injury;
- a model deceiving its developer to subvert controls.
|
- Publish a frontier AI framework.
- Machine-readable transparency summaries.
- Critical safety incident reported to the Agency and the Attorney General in 72 hours.
- Annual independent third-party audit. First.
- Public audit summary within 30 days.
- Registration before a frontier model is developed, deployed or operated.
|
Enacted; effective 1 January 2027. Framework and audit duties begin 1 January 2028. |
| MA
S.3228: Senate-engrossed frontier AI text (proposed G.L. c. 93M)
An SB 53 copy, plus a standing 180-day risk report, plus annual audits, plus independent model evaluations every 120 days (not law yet, in conference). |
- Frontier model: trained above 10^26 integer or floating-point operations.
- Large frontier developer: a frontier developer with annual gross revenues, counting affiliates, greater than $500,000,000.
|
- Catastrophic risk: not less than 50 deaths or serious injuries, or not less than $1,000,000,000 in property damage, from a single incident of:
- expert-level assistance in creating or releasing a chemical, biological, radiological or nuclear weapon;
- conduct with no meaningful human oversight that is a cyberattack, or would be the crime of murder, assault, extortion or theft;
- a model that evades the control of its developer or user.
|
- Would require: a frontier AI framework on eleven topics;
- a standing residual-risk report every 180 days;
- an annual third-party audit;
- independent evaluation of the models themselves at least every 120 days;
- critical safety incidents to the Attorney General in 15 days.
|
Not law. House non-concurred 30 July 2026; conference committees appointed in both branches. Chapter 93M would take effect 1 July 2027 if enacted. |
| CT
Public Act 26-15 (SB 5): An Act Concerning Online Safety
SB 53's whistleblower chapter only, with the penalty cut from $1,000,000 to $1,000 per violation. |
No retaliation for a safety report; an anonymous internal channel; reports escalated to officers and directors at least quarterly; notice of rights. Enacted; in force 1 October 2026; internal channel due 1 January 2027. |