New York's RAISE Act Will Require Safety Disclosures from OpenAI and Anthropic Starting in 2027

New York opened the RAISE Act registration portal on Tuesday (10th), the state's AI safety law that takes effect on January 1, 2027. Starting on that date, large frontier model developers, such as OpenAI and Anthropic, must be registered with the state, according to a report by TechCrunch.

The law specifically targets whoever trains frontier models with annual revenue above $500 million, not whoever consumes an API. In practice, this puts OpenAI, Anthropic, Google, and a handful of other labs directly in the crosshairs; a Brazilian startup that fine-tunes or builds a product on top of OpenAI's API is not subject to the obligation. But the law's text changes the type of information these labs will have to make public, and that affects any compliance team currently negotiating SLAs and due diligence with these providers.

What the RAISE Act Requires in Practice

The core obligations for frontier developers covered by the law are three:

• Publish safety frameworks on their own websites;
• Report critical safety incidents within 72 hours;
• Issue public model transparency reports.

Enforcement falls to a new body, the Office of Digital Innovation, Governance, Integrity, and Trust (DIGIT), created specifically to implement the law. Penalties reach $1 million per isolated violation and $3 million for subsequent violations, and the state can also bring an enforcement action against a company that simply fails to register on the portal. A state spokesperson has not yet detailed how these enforcement processes would work.

Why This Matters for Those Not Subject to the Law

For engineering and compliance teams in Brazil that integrate Claude or GPT into a product, the practical effect isn't a direct obligation, it's new information becoming available. A public safety framework, updated as a legal requirement, becomes a citable document in a vendor audit, in a risk report for a corporate client, or in a contract that today relies only on the lab's own marketing page.

The 72-hour deadline for reporting a critical safety incident is also relevant for whoever runs a production product dependent on these APIs: it's a transparency SLA that doesn't currently exist in any contract signed with a model provider, and now exists by force of law in at least one American state.

In short: the obligation belongs to OpenAI and Anthropic, not to the team consuming the API, but the material they'll be forced to publish becomes compliance input for any company, Brazilian ones included, that depends on these models in production and needs to justify risk to a client, auditor, or local regulator.

The Governor's Remarks and the Political Context

Governor Kathy Hochul launched the portal with remarks aimed directly at the AI lobby pushing for deregulation:

Tech moguls tell us that their vision for AI is inevitable. They tell us get on board or get left behind… That any collateral damage is worth it as long as we win the race to the AI cliff.

Kathy Hochul, Governor of New York

At the same event, Hochul inaugurated the New York Center for AI Responsibility and Research, headquartered at Binghamton University, with $55 million announced in January of this year in donations and capital. The center will independently evaluate the safety, security, and operation of advanced AI systems, partnering with DIGIT on reviewing the frameworks published by the labs.

Not the State's First AI Law

The RAISE Act joins two other New York state laws already in effect or on the way:

• A law already in effect since last year requires AI companion operators to regularly remind users that they are talking to a bot, and to make a reasonable effort to detect suicidal ideation and self-harm;
• A chatbot safety law for minors, signed by Hochul in October, restricts deceptive human-like interactions, emotionally manipulative or sycophantic responses, and encouragement of behaviors such as suicide or eating disorders. This law takes effect on June 1, 2027.

The Patchwork Anyone Serving the US Market Needs to Track

New York isn't alone. Other states have passed similar laws, with deadlines and requirements that don't line up with each other, which is starting to resemble the patchwork of state-level data privacy laws in the US.

For anyone building an AI product aimed at serving American users, or for a Brazilian company looking to open operations in the US, the practical point is that there's no single federal yardstick yet: AI safety compliance will vary state by state, with a different incident-reporting deadline in each one. It's worth tracking because the standard these states set for frontier labs tends to become, in practice, the market standard that any relevant model provider will follow globally, including for those consuming it from Brazil.