New York opens registration for large AI developers under RAISE Act
OpenAI and Anthropic must register with the state by 2027 to comply with new safety reporting and transparency rules.
The short version
- Large AI developers must register with New York's new DIGIT office by January 1, 2027.
- Companies with over $500 million in annual revenue are required to publish safety frameworks and transparency reports.
- Critical safety incidents must be reported to the state within 72 hours of occurrence.
- Violations of the RAISE Act can result in fines of up to $3 million for repeat offenses.
- A new $55 million research center at SUNY Binghamton will independently evaluate AI safety.
Why it matters
This law creates a blueprint for how states can regulate frontier AI models through mandatory transparency. While it targets large labs, it will likely change how downstream developers receive safety updates and risk information.
New York's AI safety mandate
New York has officially opened the registration portal for the Responsible AI Safety and Education (RAISE) Act. This law targets large frontier AI developers, including prominent companies like OpenAI and Anthropic. These firms have until January 1, 2027, to register with the state or face potential enforcement actions.
The RAISE Act is considered one of the most significant AI safety laws in the United States. It aims to create a structured oversight system for the most powerful models currently being developed. By requiring registration, the state can track which entities are producing high-capacity AI systems.
Governor Kathy Hochul introduced the law to ensure that technology development does not come at the expense of public safety. She expressed concern over the idea that the risks of AI development are an inevitable cost of progress. The state's goal is to prioritize community safety over corporate speed in the AI race.
Reporting and transparency rules
Companies that meet the revenue threshold of over $500 million in annual revenue face specific requirements. They must publish their safety frameworks on their company websites for public view. These frameworks will be reviewed by a newly created state office called DIGIT.
In addition to public frameworks, these companies must issue regular transparency reports regarding their models. They are also required to report any critical safety incidents to the state within 72 hours. This rapid reporting window is designed to ensure that significant failures are identified quickly by regulators.
Failure to comply with these transparency and reporting rules carries heavy financial penalties. The state can issue fines of up to $1 million for a single violation. For repeat violations, the fine amount can increase to $3 million per incident.
The role of DIGIT and SUNY
The Office of Digital Innovation, Governance, Integrity, and Trust, known as DIGIT, will serve as the state's new AI watchdog. This office is responsible for enforcing the RAISE Act and overseeing the activities of large frontier AI labs. While DIGIT has not specified the exact nature of enforcement actions, they have the authority to take action against non-compliant firms.
To support these efforts, the state is also funding a new research center at SUNY Binghamton. This center received $55 million in donations and capital to independently evaluate the security and operation of advanced AI systems. It will work alongside DIGIT to review the safety frameworks submitted by corporations.
The research center is intended to provide a neutral perspective on AI safety. By providing independent evaluations, it can offer a different set of benchmarks than those provided by the companies themselves. This could help the state determine if a model is truly safe for public use.
Impact on AI builders and users
The RAISE Act specifically targets the developers of frontier models rather than the companies that build applications on top of them. For example, a startup using an API from a large lab is not directly subject to these registration rules. However, the ripple effects will likely be felt by all developers who integrate these models into their products.
Because safety frameworks must be public, developers will have more visibility into how the major labs assess risk. The 72-hour incident reporting rule means that downstream builders might receive news of model failures much faster than before. These reports could lead to sudden API changes or new restrictions on how models can be used.
The state's move also adds to a growing patchwork of AI oversight across the country. Illinois has passed a similar law that includes independent safety audits. Together, these state-level requirements are beginning to set a de facto national standard for how frontier models are governed in the United States.
The broader regulatory landscape
The RAISE Act is not the only AI regulation currently active in New York. The state already has a companion law that requires operators to tell users when they are interacting with a bot. That law also requires companies to make reasonable efforts to detect signs of self-harm or suicidal ideation in user prompts.
Additionally, New York recently signed a chatbot safety law specifically for minors. This law restricts deceptive human-like interactions and emotionally manipulative responses from AI systems. It also prohibits AI from encouraging children to engage in dangerous behaviors.
By layering these laws together, New York is creating a comprehensive regulatory environment for different types of AI use cases. The RAISE Act focuses on the high-level safety of the models themselves, while the other laws focus on specific user interactions. This multi-layered approach aims to protect different segments of the population simultaneously.
What to watch
- Whether other states adopt New York's DIGIT model for oversight.
- How the SUNY Binghamton center's independent evaluations will be shared with the public.
- Whether the 72-hour reporting window leads to frequent API changes for downstream developers.
Try it
Check the websites of major AI labs like OpenAI or Anthropic to see if they have already published a public safety framework.
Sources and links
- ReportingOpenTools: the article we sourced
First reported by OpenTools. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.