Key Facts
- The portal opens: New York opened the registration process for its safety and transparency act on Tuesday. Large frontier AI developers such as OpenAI and Anthropic must register by January 1, 2027.
- Who is covered: Developers of large frontier models with more than $500 million in annual revenue.
- Three obligations: Publish safety frameworks on their websites, report critical safety incidents within 72 hours, and issue public model transparency reports.
- The enforcement body: The newly formed Office of Digital Innovation, Governance, Integrity, and Trust will implement the act and oversee large frontier AI labs.
- Penalties: Up to $1 million for one violation and $3 million for subsequent violations. The state can also bring enforcement actions against companies that fail to register, though a spokesperson declined to provide specifics.
- The governor’s framing: At Tuesday’s press briefing, Governor Kathy Hochul said tech moguls tell us their vision for AI is inevitable, that we should get on board or get left behind, and that any collateral damage is worth it as long as we win the race to the AI cliff.
- The independent assessor: Hochul also launched the New York Center for AI Responsibility and Research, based at State University of New York’s Binghamton University. It will independently evaluate the safety, security and operation of advanced AI systems and work with the state office to review companies’ safety frameworks. Hochul first announced the center in January with $55 million in donations and capital.
- An already-operative law: Since last year New York has enforced a law requiring AI companion operators to regularly tell users they are talking to a bot and to make reasonable efforts to detect self-harm tendencies and distress signals.
- The minors chatbot law: In October Hochul signed a separate chatbot safety law for minors restricting deceptive human-like interactions, emotionally manipulative or sycophantic responses, and encouragement of self-harm or disordered eating. It takes effect June 1, 2027.
- The peer legislation: California’s SB 53, effective since January, requires frontier labs to publish safety frameworks, enact whistleblower protections and report critical safety incidents within 15 days. Illinois’ SB 315, taking effect at the start of 2027, carries similar requirements plus independent safety compliance audits.
Competing Views
The state’s case is about symmetry and scale. Hochul describes the law as a national model for smart, sensible regulation that puts communities above corporations and people above powerful technology, and points to California and Illinois as evidence that this is a multi-state consensus rather than a single experiment.
Developers face one obligation that is qualitatively different from the others. Publishing safety frameworks and issuing transparency reports are documentation work, which an engineering organization can schedule quarterly. A 72-hour reporting window requires a standing judgment mechanism, a defined escalation path and a legal and communications review that runs continuously. That capability cannot be delivered by producing a document.
The threshold is what keeps the law from being a prohibition. Only entities with both technical scale and revenue scale are pulled in; individual researchers, academic projects and small open-source efforts are not covered, so the law places obligations on capacity rather than restricting experimentation.
Uncertainty remains about enforcement practice. The statute leaves discretion on how unregistered companies would be treated, and the state declined to describe the procedure, which means the practical bite of the law is not fully determined before it begins.
The multi-state stack is the real constraint. New York requires reporting within 72 hours, California within 15 days, and Illinois adds independent audits on top of similar duties. For a frontier lab operating across all three states, the binding standard is not any one of them but the strictest.
Panoramic Analysis
The distinction from earlier coverage deserves stating plainly. Previous reporting on this site tracked the movement in comparable legislation in California and Illinois. This is New York’s RAISE Act crossing into enforcement: the registration portal is open, the penalty schedule is set, and this is the first of the three state regimes to reach an operational stage.
The 72-hour window is the clause most likely to be underestimated. Every other obligation in the act is documentation. Safety frameworks and transparency reports can be produced on a quarterly cadence by an engineering organization, and the cost is one-time. A 72-hour incident reporting duty requires something different: within three days of an incident, the company must complete an internal determination of what happened, establish the scope of responsibility, draft an external statement and clear it through both legal and communications review. None of that machinery can be written into a document; it has to exist in the organization. A frontier lab that has never rehearsed such a notification will most likely miss the deadline the first time it matters, and the missed deadline is itself an enforceable fact.
The size of the penalties has to be read accurately, because it changes where the real constraint lies. One million and three million dollars are trivial against the revenue of a frontier lab. The law’s actual force is not the fine but the change in the structure of accountability: the state has created a dedicated enforcement office, the federal government has no comparable explicit notification duty for frontier models, and the states are now competing to set the standard. With California at fifteen days, Illinois adding independent audits and New York at seventy-two hours, the compliance baseline is effectively set by the strictest state. Any lab operating in California or New York will in practice be running to New York’s clock.
Beneath the paperwork is a shift in when liability attaches. Accountability for model companies has historically run after the fact, around harm already done, and has been settled through compensation and enforcement. This regime requires the company to build the judgment mechanism, the reporting path and the external voice before anything happens, and it extends the knowledge of a state agency. The independent assessor created this week, a center funded with $55 million and hosted inside the state university system, completes the structure by adding a third party to review the frameworks the companies publish.
There is also a time structure worth noting. Registration closes on January 1, 2027, while the chatbot law for minors takes effect on June 1 of the same year, five months later. New York is turning itself into the testing ground for tiered model regulation, and model companies face two different compliance rhythms within half a year, one addressing general frontier capability and incident reporting, the other addressing behavioral boundaries for a specific user population. The data and the precedents produced in that field will shape how any federal rule gets written.
Editor: GoodInfo Global News Team