New York’s AI safety law is now taking effect, and the registration portal is open. Frontier AI model developers with annual revenues exceeding $500 million must register by January 1, 2027, publish a safety framework on their website, report critical safety incidents within 72 hours, and release a model transparency report.
Enforcement falls to the newly established DIGIT Office. Penalties can reach $1 million per violation, and $3 million for repeat violations. Failure to register may itself be subject to enforcement, but officials have not specified exactly what would trigger it. For now, what’s clear is that developers must first register and publicly disclose their safety practices; fines will be handled during the enforcement phase.
Enforcement falls to the newly established DIGIT Office. Penalties can reach $1 million per violation, and $3 million for repeat violations. Failure to register may itself be subject to enforcement, but officials have not specified exactly what would trigger it. For now, what’s clear is that developers must first register and publicly disclose their safety practices; fines will be handled during the enforcement phase.