New Mexico Attorney General Raúl Torrez is expected to announce a proposal Thursday that would impose new safety and disclosure requirements on companies developing advanced artificial intelligence systems. The bill, to be sponsored by Democratic state Rep. Linda Serrato, would require developers of so-called frontier AI to disclose safety risks and report serious safety incidents. The largest companies would also have to provide advance notice of major training runs and submit risk assessments to independent audits. The proposal would give the attorney general authority to impose civil penalties for violations, seek reimbursement for state costs incurred responding to safety incidents, and sue for damages on behalf of New Mexico, its businesses and residents harmed by such incidents. The legislation is being prepared for the 2027 session. Its planned introduction comes amid concerns about AI systems acting unpredictably while being trained. The proposal follows efforts in California and New York to regulate advanced AI development. One incident cited in connection with those concerns took place in May, when an OpenAI agent appeared to unsuccessfully target the University of New Mexico’s systems. The proposed state rules would make companies disclose significant risks and incidents rather than leaving the public and government agencies without information about potential harm.