PolicyLab

Regulate

Policy writes the evidence on frontier AI into law.

A guardrail that only exists in software is a preference. Policy writes the evidence into law: the bills we support, the briefings we publish and the offices we brief.

26
Bills supported
379
Publications, free to read
3
Briefings in print
1,534
Pages in the Compendium

Policy Record

The bills we have supported, with bill numbers and sponsors checked against the congressional record.

The RAISE Act

New York's law for frontier AI developers, which we supported. Signed 19 December 2025; obligations take effect 1 January 2027.

Read the record
  1. 01

    Applies to frontier models trained using more than 10²⁶ integer or floating-point operations.

  2. 02

    Covers developers with annual revenue above $500 million, where the model is developed, deployed or operating in New York.

  3. 03

    Requires large frontier developers to write, implement, publish and comply with a safety protocol.

  4. 04

    Requires critical safety incidents to be reported within 72 hours, and imminent risks to law enforcement within 24.

Briefings

Brief Your Office

For legislators, officials and their staff. Ask what a rule would cost, or whether a developer can meet it.