United States of America: Responsible AI Safety and Education (RAISE) Act, including transparency obligations enters into force with grace period

On 19 March 2026, the Responsible Artificial Intelligence (AI) Safety and Education (RAISE) Act entered into force with a grace period, introducing obligations for developers of frontier AI models. It imposes transparency obligations requiring all frontier developers to publish a report before or concurrently with deploying a new or substantially modified model, covering its intended uses, supported languages, output modalities, and applicable restrictions. It further requires large frontier developers, those with annual gross revenues exceeding USD 500 million, additionally required to publish summaries of catastrophic risk assessments and third-party evaluator involvement. Large frontier developers must also write, implement, and publicly publish a frontier AI framework setting out documented protocols for identifying, assessing, and mitigating catastrophic risks, which must be reviewed at least annually and updated within thirty days of any material modification. Further, no large frontier developer may operate in New York State without filing a current disclosure statement with the office, identifying the developer, its ownership structure, and designated points of contact, renewed every two years or upon any material change, with failure to comply attracting civil penalties of USD 1'000 per day.