OpenAI, Anthropic face new oversight as New York’s AI safety law takes effect

New York opened a registration portal on Tuesday for the RAISE Act, its AI safety and transparency law, which takes effect at the start of 2027. Large frontier AI developers — defined in the law as companies with more than $500 million in annual revenue — will be required to register through the portal by January 1, 2027. That definition covers labs such as OpenAI and Anthropic, but not smaller startups or most open-source model publishers.
The obligations extend beyond registration. Covered companies must publish safety frameworks on their own websites, report critical safety incidents within 72 hours, and issue public model transparency reports. Enforcement sits with a newly formed state body, the Office of Digital Innovation, Governance, Integrity, and Trust (DIGIT), which will administer the law and oversee the large frontier labs. Penalties reach as high as $1 million for a single violation and $3 million for subsequent ones. The state can also bring enforcement actions against companies that fail to register at all, though a spokesperson declined to describe those measures in detail.
Governor Kathy Hochul framed the effort at a Tuesday press briefing around skepticism of industry assurances, saying tech leaders describe their AI vision as inevitable and that "any collateral damage is worth it as long as we win the race to the AI cliff." She described the law as "a national model for smart, sensible AI regulation, a model that puts communities above corporations and people above powerful technology."
Hochul also launched the New York Center for AI Responsibility and Research, based at the State University of New York's Binghamton University. The center is intended to independently evaluate the safety, security and operation of advanced AI systems and to work with DIGIT in reviewing companies' safety frameworks. Hochul first announced the center in January with $55 million in donations and capital.
For developers, the 72-hour incident reporting clock is the provision most likely to change day-to-day operations. A critical incident under this regime implies a deadline tight enough that labs will need pre-built escalation paths, on-call review and documentation templates ready before an event occurs, rather than assembling a report after the fact. The requirement to publish a safety framework and periodic transparency reports also creates public artifacts that competitors, researchers and regulators can compare across labs — a dynamic that tends to push disclosure norms upward once one major developer sets a format.
The $500 million revenue threshold is worth noting as a design choice. It concentrates obligations on a handful of frontier labs while leaving smaller model developers outside the registration regime, which is roughly the same frontier-lab framing California used in SB 53. A likely trade-off is that the law reaches the largest players quickly but leaves gaps for well-funded firms below the line, or for companies that distribute powerful models through partners rather than selling them directly.
New York's other AI statutes show a pattern of layered, narrowly targeted rules rather than one comprehensive code. The state last year began enforcing a law requiring AI companion operators to regularly tell users they are talking to a bot and to make reasonable efforts to detect suicidal ideation and self-harm. In October, Hochul signed a separate chatbot safety law for minors, restricting deceptive human-like interactions, emotionally manipulative or sycophantic responses, and encouragement of behaviors such as suicide or disordered eating; that law takes effect June 1, 2027.
Other states are moving in parallel. California's SB 53 took effect in January and requires frontier labs to maintain and publish safety frameworks, provide whistleblower protections, and report critical incidents within 15 days. Illinois's SB 315 takes effect at the start of 2027 and carries similar requirements plus independent safety compliance audits. The spread of these regimes — with different reporting windows, such as New York's 72 hours against California's 15 days — suggests covered labs will end up managing state-by-state compliance calendars rather than a single national standard.
Why it matters: Frontier labs above the revenue threshold now face concrete, dated obligations in the largest US state by population, with registration due by January 1, 2027 and fines reaching $3 million for repeat violations. Developers at covered companies should expect compliance work around incident response, public safety documentation and transparency reporting to become routine engineering and legal overhead. Teams building below the threshold are not directly captured, though in practice larger partners may push similar disclosure expectations down the supply chain — an inference, not a stated requirement.
Based on reporting from the original publisher. Visit the source for full context and later updates.
Publisher excerpt
New York is preparing to enforce its sweeping AI safety law, requiring major developers like OpenAI and Anthropic to register, disclose safety practices, and report critical incidents or face fines.