New York opened a state portal on Tuesday requiring large frontier artificial intelligence developers to register for oversight, as Governor Kathy Hochul launched an independent evaluation programme. Under the New York AI law, companies face fines of up to $3 million if they fail to disclose critical safety incidents, publish safety frameworks, or issue model transparency reports.
How the registration rules work
The registration requirement forms the first operational stage of the state’s RAISE Act, which takes full effect on 1 January 2027. It targets frontier developers with more than $500 million in annual revenue, placing leading artificial intelligence creators such as OpenAI and Anthropic under formal state oversight.
Albany’s newly established Office of Digital Innovation, Governance, Integrity, and Trust, known as DIGIT, will run the registry and police compliance, TechCrunch reported. The office holds the authority to penalise non-compliant firms, with fines set at $1 million for a first breach and rising to $3 million for subsequent infractions.
State regulators can also take enforcement action against companies that refuse to register through the portal. A spokesperson for the state declined to specify what measures officials would take if a frontier laboratory refused to comply before the 2027 deadline.
Independent testing backed by Empire AI
Alongside the registration portal, Hochul inaugurated the New York Center for AI Responsibility and Research at the State University of New York at Binghamton. The governor initially revealed plans for the facility in January, funding it with $55 million in capital grants and private donations.
The Binghamton facility will assess the safety, security, and operation of commercial models, working directly alongside DIGIT to review companies’ published safety frameworks. More than 80 researchers across New York will conduct the technical assessments, Engadget reported.
Testing will rely on computing infrastructure supplied by Empire AI, the state’s $500 million public-private supercomputing initiative. Researchers will track system hazards, evaluate how systems function, and study safe applications in areas such as clinical medicine and scientific discovery, supplying findings directly to state officials.
Hochul defended the decision to build public evaluation capacity rather than relying on internal corporate audits. “The stakes are too high to simply take AI companies at their word,” Hochul said in a statement announcing the programme.
The state testing setup faces one clear structural boundary. Binghamton researchers can only examine models and code that developers release publicly. Unreleased models and internal pre-deployment tests remain outside state visibility, though the RAISE Act seeks to narrow that gap by compelling firms to release detailed model transparency logs.
Mandatory incident reports within 72 hours
Under the RAISE Act, commercial developers meeting the revenue threshold must publish detailed safety frameworks directly on their websites. The statute also creates strict notification rules, compelling companies to report any critical safety incidents to state regulators within 72 hours.
The mandatory reporting timeline gives New York one of the tightest safety incident notification windows in the United States. During a Tuesday press briefing, Hochul argued that state intervention had become essential to check the pace of commercial model development.
“Tech moguls tell us that their vision for AI is inevitable,” Hochul said during the briefing, criticising the tech sector’s arguments that collateral damage must be accepted to maintain competitive speed.
State standards outpace federal rules
The frontier model rules expand an existing web of algorithmic legislation in New York. Last year, the state started enforcing companion bot legislation that forces bot operators to remind users that conversations are automated and requires systems to detect indicators of self-harm and suicide.
In October, Hochul signed a separate chatbot safety law focused on children. Taking effect on 1 June 2027, that statute bars deceptive human mimicry, emotionally manipulative interactions, sycophantic conversational patterns, and outputs encouraging eating disorders or self-harm.
Other states are moving on similar schedules. California’s SB 53 took effect in January, forcing frontier labs to post safety blueprints, protect corporate whistleblowers, and report critical safety failures within 15 days. Illinois passed SB 315, which begins enforcement at the start of 2027 with similar disclosure rules and independent compliance audits.
Frontier AI laboratories now face a firm timeline. Covered developers must register through the state portal and prepare their transparency documentation before the RAISE Act comes into force on 1 January 2027.