PRIA Logo™
New YorkS 88282025-2026 Regular SessionSenate

Relates to transparency and safety requirements for developers of artificial intelligence models

Sponsored By: Andrew Gounardes (Democratic)

Signed by Governor

RULESWAYS AND MEANS

Personalized for You

How does this bill affect your finances?

Personalize government policy and PRIA will tell you what this bill means for your household, plus every other piece of legislation we track. PRIA reads each provision against your financial profile to show you exactly what matters to your wallet.

Free to download

Bill Overview

Analyzed Economic Effects

8 provisions identified: 0 benefits, 2 costs, 6 mixed.

AI developers face fines for violations

The Attorney General can sue large frontier developers for listed safety and transparency violations. These include missing required documents, prohibited statements, unreported incidents, and failure to follow their own safety plans. Penalties can reach $1 million for a first violation and $3 million for each later violation, based on severity. After notice and a hearing, the office can impose $1,000 per day for missing disclosures or uncorrected false information. Specified filing or payment violations can also bring a penalty equal to assessments owed.

Large AI developers file and pay

Large frontier developers must file current disclosures and pay their required share before developing, deploying, or operating covered models in New York. Filings include business names, the main address, every New York office address, ownership details, and three disclosure contacts. Private or closely held entities must identify owners benefiting from at least 5% of equity, including former qualifying owners from the past five years; publicly traded ultimate parents use the specified 50% threshold. Filings must be renewed every two years, or sooner after a model ownership transfer or major reported change; the office publishes a filer list without specified contact details. The Department of Financial Services charges developers proportional shares of direct and indirect program costs, with no fixed dollar amount stated.

AI developers face fast incident deadlines

New York's financial services office must provide an incident-reporting channel for frontier developers and the public. Reports give the incident date, description, why it qualifies, and whether it involved internal model use. Critical incidents include stolen or altered model weights, severe-risk harms, loss of control causing death or injury, and deception to defeat monitoring. Developers must report within 72 hours of confirming an incident or learning facts supporting a reasonable belief it happened; imminent death or serious injury requires notice to an appropriate authority within 24 hours. Developers may use state-approved, equivalent or stricter federal reporting standards after notifying the office, but must send federal-required reports to the office at the same time; approval can be revoked.

AI law limits private lawsuit rights

Individuals and organizations cannot use this law itself as the basis for a private lawsuit over violations. Other legal duties still apply, and rights and remedies under other laws remain available. A drop in the value of an ownership stake does not count as property damage or loss under this law. Defendants may still argue that other parties or factors caused harms linked to severe AI risks.

AI releases require truthful public reports

Frontier developers must publish reports before or when they deploy new or substantially changed models. Reports list release dates, supported languages and output types, intended uses, limits, and developer contacts. Large developers must also summarize severe-risk assessments, results, outside evaluators, and other safety-plan actions. Developers cannot make materially false or misleading risk claims; large developers also cannot misrepresent safety-plan compliance, except for good-faith, reasonable statements. Developers may hide details only as needed for trade secrets, cybersecurity, public safety, national security, or legal compliance; they must explain redactions when permitted and keep complete records for five years.

Large AI developers must follow safety plans

Large frontier AI developers must write, publish, carry out, and follow safety plans. Plans must cover severe-risk reviews, deployment choices, outside assessments, cybersecurity, incident response, oversight, and internal model use. Covered risks include expert help creating weapons, unsupervised cyberattacks or violent crimes, and evading developer or user controls, with specified exclusions. Developers must review plans at least yearly and publish major changes with reasons within 30 days. New York's financial services office can make rules and consider extra reports, sharing plans, protocols, or safety-plan submissions; these extras are not automatically required.

New York sets powerful AI coverage

The law sets January 1, 2027, as its main effective date, subject to coordination with related 2025 legislation. It covers frontier AI models developed, deployed, or operating partly or fully in New York. These are general-purpose models trained using more than 10^26 floating-point operations, including training and later changes. Large developers are those whose combined revenue with affiliates exceeded $500 million in the prior calendar year. Accredited New York colleges and universities are exempt while doing academic AI research; the named Empire AI consortium and institute are also exempt.

State reviews AI risks with confidentiality

Large frontier developers must send summaries of internal-use severe-risk assessments every three months, unless another reasonable schedule is agreed in writing. The office must provide confidential submissions and take reasonable steps to limit access. These assessments and critical incident reports are exempt from New York's public records law, including copies sent to other agencies. The office may review public reports and share reports with government agencies based on severity, ongoing risk, legal needs, or coordination. Beginning January 1, 2028, it must publish yearly anonymous, combined reports and recommend updates, excluding protected or legally restricted information.

Sponsors & Cosponsors

Sponsor

  • Andrew Gounardes

    Democratic • Senate

Cosponsors

There are no cosponsors for this bill.

Roll Call Votes

All Roll Calls

Yes: 217 • No: 3

House vote • 3/11/2026

Assembly floor vote

Yes: 140 • No: 2

Senate vote • 1/28/2026

FLOOR Vote

Yes: 58 • No: 1

committee vote • 1/20/2026

Rules Committee Vote

Yes: 19 • No: 0

Actions Timeline

  1. SIGNED CHAP.96

    3/27/2026Senate

  2. DELIVERED TO GOVERNOR

    3/20/2026Senate

  3. RETURNED TO SENATE

    3/11/2026House

  4. PASSED ASSEMBLY

    3/11/2026House

  5. ORDERED TO THIRD READING RULES CAL.85

    2/25/2026House

  6. SUBSTITUTED FOR A9449

    2/25/2026House

  7. REFERRED TO WAYS AND MEANS

    1/28/2026House

  8. DELIVERED TO ASSEMBLY

    1/28/2026Senate

  9. PASSED SENATE

    1/28/2026Senate

  10. ORDERED TO THIRD READING CAL.94

    1/20/2026Senate

  11. REFERRED TO RULES

    1/8/2026Senate

Bill Text

  • Original

    1/8/2026

    PDF

Related Bills

Back to State Legislation