New York RAISE Act — frontier model evaluation.
How Hael produces the frontier-model risk evaluation, safety testing documentation and incident-response protocols the New York RAISE Act requires — sealed for the State.
What the RAISE Act requires
The New York Responsible AI Safety and Education (RAISE) Act establishes obligations for developers of frontier AI models with material exposure to New York. Covered developers must conduct documented safety evaluations before deploying frontier models, contract independent third-party safety testing for highest-risk capabilities, and maintain incident-response protocols capable of containing or rolling back deployments where critical safety risks materialise.
The statute reflects New York's posture on frontier AI risk and complements California SB 53. Where SB 53 emphasises published transparency, RAISE emphasises substantive safety evaluation and operational response capability. Documentation is the substantive artefact: pre-deployment safety evaluations, third-party testing reports, incident-response playbooks and post-incident review records must be maintained and producible on regulator request.
How Hael runs it
Hael runs the safety evaluation as part of the lifecycle state machine. Frontier model deployments require evaluation gate completion before lifecycle progression. Evaluation evidence — methodology, test sets, results, mitigations — is collected continuously and sealed in the audit chain. Third-party testing partners are wired in with scoped access to evaluation infrastructure; their reports are sealed alongside internal evaluations.
Incident-response protocols are operationalised in the platform's incident chain. Containment and rollback actions are pre-authorised against deployment registry state. Post-incident review records are generated automatically and sealed with hash-chained provenance, ready for regulator engagement.
Questions
Which developers does the RAISE Act apply to?
Frontier AI model developers with material exposure to New York — meaning models trained using compute above the statutory threshold and either developed in New York, deployed to New York users at material scale, or processing New York personal information at material scale.
What does 'independent' third-party testing mean?
The testing partner must have no equity interest in the developer, no employment relationship with the developer in the previous two years, no compensation structure dependent on testing outcomes, and operational independence from the developer's evaluation team. Documentation of independence is a regulator-facing artefact.
How does RAISE interact with California SB 53?
Frontier developers with material exposure to both states must comply with both. SB 53 emphasises published transparency; RAISE emphasises substantive evaluation and incident response. The same evidence base produces both artefacts — the published safety framework for SB 53 and the documented evaluations and response protocols for RAISE — with no duplicate collection.
See Hael run your RAISE Act obligations.
Safety evaluation infrastructure, third-party testing access, incident-response protocols — wired against your frontier model operations.