Law / United States / 
New York
Responsible AI Safety and Education Act (RAISE Act)
N.Y. Gen. Bus. Law art. 44-B (§§ 1420-1425), ch. 699 of 2025
A citation is an address, not a summary. The first part names the law; what follows narrows it to the exact section, article or paragraph.
What follows is LexLint's own research summary of this law, not legal advice.
In force 6 months, effective 19 March 2026.
An AI governance rule binding private bodies.
As of 15 September 2026.
What it requires
- If you are a large developer of a frontier AI model, write a safety and security protocol addressing severe risks and conspicuously publish it, with appropriate redactions, transmitting an unredacted copy to the Attorney General on request.
- Disclose each safety incident affecting a frontier model to the Attorney General within 72 hours of learning of it.
- Review your safety and security protocol at least once a year, and if that review materially modifies it, publish the modified protocol and transmit it to the Attorney General and the Division of Homeland Security and Emergency Services the same way you published the original
If you get it wrong
Private right of actionNo
Penalty structure
Civil penalty of up to $10,000,000 for a first violation and up to $30,000,000 for any subsequent violation of the safety and security protocol duty (section 1421).
- Rule
- Per violation only
- As of
- 6 September 2026
- Currency
- USD
- Per violation unit
- Violation
- Per violation amount
- 10,000,000
Who enforces it
Enforcement body
New York Attorney General
What it reaches
How the hook was established
express
What makes it apply
Operator establishment, Market targeting, Place of effect
What it makes you log
Who may demand the log
Regulator
What the log must hold
Decision basis
Log retention
The stated period is the length of the model's deployment plus five years, so the deployment portion has no fixed length; five years is the fixed component the statute names and is recorded here as the minimum.
- Unit
- Years
- As of
- 21 September 2026
- Basis
- Purpose bound
- Minimum value
- 5
Logging duty
A large developer must record, as and when reasonably possible, the specific tests and test results behind any assessment of a frontier model that this section or its own safety and security protocol requires, in enough detail for a third party to replicate the testing procedure, and retain that record for as long as the model is deployed plus five years. The same subdivision requires retaining an unredacted copy of the protocol itself, including records and dates of updates, for the same period; the duty applies to every frontier model a large developer deploys, not only one found to pose an unreasonable risk.
- Kind
- Explicit
- As of
- 21 September 2026
- Provision
- § 1421(1)(d)
- Trigger
- all_systems
Why the legislator wanted it
- As of
- 21 September 2026
- Quote
- Disclose major security incidents, so that no one has to make the same mistake twice.
- Source link
- https://www.nysenate.gov/legislation/bills/2025/S6953/amendment/B
- Source kind
- Explanatory memorandum
- Purpose
- incident_reconstruction
- oversight_and_correction
Who checks it
Audit expectation
periodic
Who audits it
Self
Where the report goes
Published, Filed with regulator
What this law does
Drafted with AI from the cited sources under the direction of UnGovr staff. UnGovr holds editorial responsibility for this page.
A large developer of a frontier AI model must implement a written safety and security protocol before deploying the model, retain an unredacted copy of it for as long as the model is deployed plus five years, conspicuously publish a redacted copy and transmit that copy to the Attorney General and the Division of Homeland Security and Emergency Services, and grant either of them access to the protocol on request.
It must record the specific tests and test results behind any assessment the section or its own protocol requires, in enough detail for a third party to replicate the testing procedure, and retain those for the same period.
It must not deploy a frontier model that would create an unreasonable risk of critical harm, must review its protocol annually and republish it if that review makes a material modification, must disclose each safety incident to the Attorney General and the Division within 72 hours, and must not knowingly make false or materially misleading statements in the documents the section produces.
The Attorney General may bring a civil action for a civil penalty of up to $10,000,000 for a first violation and $30,000,000 for any subsequent violation, or for injunctive or declaratory relief; the act creates no private right of action. Signed December 19, 2025 as chapter 699 of the Laws of 2025, and in force since March 19, 2026, the ninetieth day after enactment.
Note that the enacted print, S6953-B, carries no third-party audit: the annual independent audit of compliance, its report contents and its publication duty appeared in the original print and in S6953-A and were dropped before passage, and the word audit does not appear in the chaptered text.
When LexLint raises it
trains_models
Read the law
New York State Senate, bill text and enactment history for S6953-B (2025)
Every line above is drawn from the primary source linked here, read on the date shown. This is a research summary, not legal advice.