New York Seeks to RAISE the Bar on AI Regulation

A blog post reports that New York lawmakers approved the RAISE Act, short for the Responsible AI Safety and Education Act, on June 12, 2025. The measure then went to the governor for a final decision. The post sees it as another state-level effort to regulate AI, one that could gather pace now that a ten-year moratorium on AI regulation was dropped from the One Big Beautiful Bill.

The post says the act draws on California's SB 1047, which Governor Gavin Newsom vetoed in September 2024 over innovation worries. It targets frontier models and tries to prevent critical harm, meaning more than 100 deaths or injuries, or over $1 billion in damage to money or property, tied to chemical, biological, radiological or nuclear weapons or to largely autonomous model conduct that would be a serious crime if a person did it. Coverage is limited to developers that spent more than $100 million on computing to train a model, or more than $5 million where a smaller model was trained on a larger one and has similar capabilities to that larger model, and that offer the model to New Yorkers.

Before deployment, a covered company must write a safety and security protocol, keep an unredacted copy for the length of deployment plus five years, post it publicly and send it to the state's homeland security division, and log the tests it ran and the results. The attorney general can seek civil fines reaching $10 million for a first violation and $30 million after that, including for skipping the yearly independent audit. Appropriate redactions are allowed for safety, trade secrets, legal confidentiality or privacy. The post warns that record-keeping costs could squeeze out smaller firms and calls the bill's fate uncertain.