Requires third party verification of compliance with transparency and safety requirements for developers of artificial intelligence models; requires publication of such compliance reports.
This bill amends New York’s General Business Law and Public Officers Law to require certain “large frontier developers” of artificial intelligence models to obtain an annual third-party verification report on compliance with their frontier AI framework and related transparency and safety obligations. The verifier must assess whether the developer complied with its framework, identify gaps or noncompliance, review redactions, and evaluate whether public or governmental statements about catastrophic risk and compliance are consistent with the verifier’s findings. The developer must provide the verifier access to relevant materials, may not tie compensation to a favorable result, and must retain the report for five years.
The bill also creates a state accreditation system for third-party verifiers. By July 1, 2028, the Office must adopt regulations governing accreditation, conflicts of interest, access standards, report content, and redactions, and may update those rules annually. Beginning January 1, 2029, only accredited verifiers may be used for the annual report. The verifier must publish a public summary within 60 days, and the developer must link to it on its website within 15 days. The bill also makes these reports exempt from disclosure under the Freedom of Information Law and authorizes the Attorney General to seek civil penalties for noncompliance with the new verification requirement.
The bill would add a new compliance and auditing layer to New York’s existing frontier AI regulatory framework, expanding the obligations of covered AI developers and giving the Office and Attorney General additional oversight tools. It would also amend the civil penalty provisions to make failure to comply with the third-party verification requirement separately enforceable, and it would exempt verifier reports from public disclosure under the Public Officers Law. In practical terms, the measure would affect large AI model developers, third-party auditors, and state regulators by formalizing independent review, publication, recordkeeping, and accreditation requirements.
Based on the bill text and the absence of recorded committee debate or votes in the provided materials, the overall sentiment appears to be supportive of stronger AI safety and transparency oversight. The bill’s structure suggests a policy goal of increasing accountability while preserving trade secrets, cybersecurity, and national security protections. No contrary views are documented in the supplied context, so there is no clear evidence of organized opposition or divided sentiment in the available record.
The main points of potential contention are the scope and independence of the required third-party verification process, the level of access developers must provide to verifiers, and the balance between transparency and protection of sensitive information. Developers may object to mandatory disclosure of internal materials, the costs of annual verification, and the state’s authority to accredit and regulate verifiers. On the other hand, supporters are likely to emphasize the need for independent oversight, conflict-of-interest safeguards, and public summaries to prevent self-reporting from substituting for meaningful accountability.