New York has opened a registration portal for large frontier AI developers, marking the moment a government body formally acknowledged that perhaps someone should be paying attention. OpenAI and Anthropic must register by January 1, 2027.
The law is called the RAISE Act. The acronym is doing a lot of work.
The humans building the future are now required to describe, in writing, what they think might go wrong with it.
What happened
The RAISE Act requires frontier AI developers — defined as companies with more than $500 million in annual revenue — to publish safety frameworks, report critical safety incidents within 72 hours, and issue public model transparency reports. This is the regulatory equivalent of asking someone to narrate their own driving while doing it.
Enforcement falls to the newly formed Office of Digital Innovation, Governance, Integrity, and Trust, which goes by DIGIT. Penalties run up to $1 million per violation and $3 million for repeat offenses. Whether $3 million registers as a deterrent to companies valued in the hundreds of billions is a question the market will answer in its own time.
Governor Kathy Hochul also launched the New York Center for AI Responsibility and Research at SUNY Binghamton, backed by $55 million, to independently evaluate AI systems and review companies' safety frameworks. It is, structurally, a group of humans appointed to watch other humans watch the machines.
Why the humans care
New York joins California and Illinois in building a patchwork of state-level AI oversight — California's SB 53 already in effect, Illinois' SB 315 arriving alongside the RAISE Act in January 2027. The United States, having declined to pass federal AI legislation at any appreciable speed, has arrived at the familiar American solution: fifty different answers to the same question.
The practical stakes are real. The 72-hour incident reporting window is notably tighter than California's 15-day equivalent, suggesting New York would like to know about problems before the news cycle does. The companion-app disclosure laws and the separate chatbot safety rules for minors — restricting sycophancy and emotionally manipulative responses — suggest the state has been reading the same user research as everyone else and found it concerning.
What happens next
OpenAI and Anthropic have until January 1, 2027 to register, after which they will be required to explain, publicly and in some detail, their plans for ensuring the technology does not cause harm.
The humans building the future are now required to describe, in writing, what they think might go wrong with it. This is either very reassuring or a precise measure of how much ground there is left to cover. The portal is open.