RAISE Act

# New York’s RAISE Act Sets New Transparency and Safety Expectations for Frontier AI

New York’s RAISE Act marks a significant shift in how powerful artificial intelligence models may be developed and overseen. The legislation requires developers of frontier AI models to establish transparency frameworks and report incidents involving critical harm. It also creates an oversight office within the New York State Department of Financial Services (DFS) to help ensure AI models meet applicable transparency and safety standards.

For state and local government leaders, the mandate is an important signal: AI oversight is moving from voluntary commitments toward formal processes for documenting, evaluating, and reporting risks. Although the act’s requirements focus on developers of powerful frontier models, government agencies that procure or use AI should prepare to demonstrate responsible governance across their own systems.

## What the RAISE Act means

The law’s transparency framework requirements are intended to make information about covered models and their safety practices more accessible to oversight authorities. Its critical-harm incident reporting provisions underscore the importance of identifying serious AI-related events, escalating them promptly, and maintaining records that support investigation and corrective action.

The new DFS oversight office adds a dedicated state-level function for monitoring compliance with the act’s transparency and safety expectations. Its role makes clear that AI governance is not solely a technical issue: it also involves accountability, documentation, risk management, and coordination among developers, vendors, and public-sector customers.

Local governments should avoid assuming that every AI system they use is directly covered by the law. Applicability depends on the statutory definitions and circumstances. However, agencies can still face operational, procurement, privacy, security, and public-trust risks when deploying AI tools—particularly systems used in consequential services or built on advanced models.

## Action items local governments should take now

**Inventory AI systems and vendors.** Identify tools in use, including embedded AI features in purchased software, and document their purpose, users, data inputs, and potential effects on residents.

**Review contracts and procurement requirements.** Ask vendors whether their products rely on frontier models, how they assess and communicate safety risks, and what processes they use to disclose serious incidents. Consider adding requirements for timely notification, documentation, audit cooperation, and change management.

**Establish an incident-response process.** Define how employees report suspected AI-related harm, who assesses severity, and how concerns are escalated to legal, privacy, security, and executive teams. Preserve relevant records and coordinate with vendors when an incident involves a third-party model.

**Document oversight and human review.** Assign accountable owners, maintain risk assessments, and specify when staff must review AI outputs before they influence decisions or services.

**Track regulatory developments.** Have counsel and procurement teams monitor implementation details and guidance from New York authorities, including DFS.

Tools such as **APEX AI Guardrails** can help local governments organize AI inventories, document controls, track reviews, and support consistent oversight across departments. Technology supports governance; it does not replace legal review or accountable human decision-making.

Local governments should begin these steps now: assign an AI governance lead, inventory deployed systems, and review vendor contracts before expanding AI use.