WeVote
Log In

Bill

Bill

S 10456

Relates to minimum standards for large frontier developer frontier AI frameworks

2025 Regular Session Introduced by Andrew Gounardes

New York would require large frontier AI developers to meet minimum safety standards, with the state Office drafting and annually updating regulations guided by best practices and

REFERRED TO INTERNET AND TECHNOLOGY
0
WeVote Research Nonpartisan
Bill Summary · S 10456

Overview

S.10456 (2025-2026) from New York Senate proposes establishing minimum standards for large frontier AI frameworks to mitigate catastrophic risk. The bill would require the state Office to adopt regulations setting these standards and to periodically review and update them.

Purpose and intent

  • Establish minimum standards for “large frontier developers’ frontier AI frameworks” to prevent unreasonable levels of catastrophic risk.
  • Create a regulatory framework that incorporates best practices and aligns with national/international standards and other jurisdictions.
  • Ensure ongoing vigilance through annual reviews and updates in light of scientific, technical, and incident-driven developments.

Key provisions

  • Section 1430 (New General Business Law):
    • The Office (presumably the state Office relevant to AI regulation) must adopt regulations by July 1, 2028, to require large frontier AI frameworks to meet minimum standards addressing catastrophic risk.
    • Regulation development should consider:
    • Industry best practices
    • National and international standards
    • Laws/regulations in other states/jurisdictions
    • The Office must solicit feedback from:
    • Academia
    • Empire AI Research Institute (established under existing Economic Development Law)
    • Attorney General
    • Civil society
    • Frontier model developers
    • Foundation model developers
    • The Office must review the regulations annually and update as needed to reflect:
    • Advances in scientific understanding of catastrophic risk
    • Technical progress in frontier models
    • Critical safety incidents
    • Identified regulatory gaps

Scope and affected parties

  • Affects “large frontier developers” and their frontier AI frameworks (implied to include frontier models and possibly related foundation models).
  • Affects entities involved in AI development and deployment, including industry players, researchers, and regulators.
  • Involves multiple state stakeholders in its development and ongoing oversight (academia, AG, civil society, industry researchers).

Procedural and timeline details

  • Regulatory action:
    • Final regulations due by July 1, 2028.
    • Regulations to be created under Article 2 of the State Administrative Procedure Act.
  • Review cadence:
    • Annual review by the Office of the regulations enacted under this section.
    • Potential updates triggered by new scientific findings, model advancements, safety incidents, or regulatory gaps.
  • Effective date:
    • The act becomes law on the thirtieth day after enactment.
    • If Article 44-B of the General Business Law (as added by chapter 96 of 2026) has not taken effect by that date, the act becomes effective 30 days after that article takes effect.

Potential impact

  • Creates a formal, state-level standard-setting process for high-risk AI frameworks.
  • Promotes risk mitigation practices in AI development and deployment.
  • Encourages cross-stakeholder engagement in regulatory design.
  • Establishes a pathway for ongoing updates as technology and risk understanding evolve.
  • The exact standards (e.g., safety, transparency, auditing, interoperability requirements) will be defined in the regulations adopted by the Office and are not specified in the bill text.

Summary in plain language

New York seeks to require large frontier AI developers to meet minimum safety standards for their frontier AI systems. The state Office would draft these standards, guided by best practices and international norms, and would gather input from researchers, industry, government, and civil society. The rules would be reviewed annually and updated as technology and safety knowledge advance. Final regulations are due by mid-2028, with the bill taking effect about a month after enactment, subject to related legislation.

Compiled from official sources — confirm details with the bill’s official record.

Sign in to ask a question.