New York enacts RAISE Act for frontier AI: stricter rules

New York Governor Kathy Hochul signed the Responsible AI Safety and Education (RAISE) Act on December 19, 2025. The law imposes reporting, safety and transparency requirements on developers of high-capability or “frontier” AI models. The governor’s office framed it as a state-level framework; Axios and Reuters reported the package targets model-risk reporting, safety protocols and public disclosures. citeturn0search0turn0search1

The law requires developers to document safety practices and to disclose models that meet New York’s “frontier” criteria as defined in the bill. Reuters and Axios note it builds on earlier state efforts such as California’s AI rules. The Verge reported that advocacy and higher-education groups secured narrower language during negotiations and characterized some final changes as weakening the original draft. citeturn0search2turn0search3

For model developers and platform operators, the RAISE Act raises concrete compliance tasks: produce reproducible risk assessments, expand logging and transparency about capabilities, and establish incident-reporting workflows. Expect legal, engineering and safety teams to coordinate on documentation, third-party audits and deployment gating for models that meet New York’s frontier thresholds. citeturn0search0turn0search1

Why It Matters

• Designate compliance leads: have engineering, safety and legal map models to New York’s "frontier" criteria and document controls.
• Prepare audit-ready artifacts: reproducible risk assessments, consistent logging, and retained artifacts are required for reporting and third-party review.
• Operationalize incident reporting: implement monitoring, escalation playbooks and timelines to meet mandated disclosure and response obligations.
• Apply deployment gating: restrict model access or rollouts in New York until compliance artifacts, tests and any required audits are complete.