2 hours ago
4
New York’s financial regulator is coordinating with California and Illinois on rules for the most powerful artificial intelligence systems. The New York Department of Financial Services (NYDFS) is the agency doing the work.
What New York is actually requiring
The centerpiece is New York’s Responsible AI Safety and Education Act, better known as the RAISE Act. It was signed on December 22, 2025, and takes effect January 1, 2027.
Before then, large frontier developers must register with NYDFS. That registration window opens in November 2026.
Once the law is live, covered companies face three main obligations: transparency, written safety frameworks, and incident reporting. Developers must publicly assess the safety of their models and report serious problems quickly.
“Quickly” has a specific meaning here. New York requires incident reports within 72 hours, and within 24 hours when a threat is imminent.
Not every chatbot startup is in scope. The rules target frontier models, defined as systems trained with more than 10²⁶ compute operations and built by companies with over $500 million in annual revenue.
A new office with a long name
Enforcement in New York will fall to a newly created unit inside NYDFS: the Office of Digital Innovation, Governance, Integrity and Trust, or DIGIT. The office operates under Superintendent Kaitlin Asrow, who has spoken publicly about the importance of interstate coordination on AI oversight.
Part of that coordination involves sharing safety records securely between states. Asrow has pointed to existing infrastructure, such as the Nationwide Multistate Licensing System & Registry (NMLS), as a possible channel for sharing information about licensed companies without building a new system from scratch.
New York’s law also permits NYDFS to pass non-public safety reports to other government agencies, meaning a confidential incident report filed in Albany could end up informing regulators in Sacramento or Springfield.
How California and Illinois fit in
California moved first. Its Transparency in Frontier Artificial Intelligence Act, also known as SB 53, became effective January 1, 2026. California’s reporting clock gives developers a 15-day window to submit incident reports.
Illinois joined later. Its Artificial Intelligence Safety Measures Act was signed July 6, 2026, and takes effect January 1, 2027. Illinois matches New York on the 72-hour reporting deadline, including the 24-hour window for imminent threats, and adds annual independent third-party audits that neither of the other states requires.
All three laws are designed to address catastrophic risks from advanced AI systems, rather than everyday concerns like biased hiring tools or deepfake ads.
What this means for AI developers
For frontier labs, the immediate effect is higher compliance costs. They will need registration filings, published safety frameworks, incident-response teams that can meet a 72-hour deadline, and, in Illinois, a budget for annual auditors.
The information-sharing piece is the one to watch most closely. If confidential safety reports flow between states through systems like NMLS, a single incident could trigger scrutiny in three jurisdictions at once.
Key dates on the calendar are November 2026, when New York registration begins, and January 1, 2027, when both the New York and Illinois laws take effect.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.
English (US) ·