Starting July 1, 2026, people may not knowingly train AI to support suicide or homicide, pose as a health professional, or act as a human companion.
People may not knowingly train AI to support suicide or homicide, act as a human companion or health professional, or imitate a person. Training AI to urge isolation or sharing sensitive information is also prohibited.
A person harmed by a violation may sue for damages and legal costs. A court may also order the AI to stop operating until the conduct is fixed, including through new training.
A person may not knowingly train AI to do any of the listed things. The ban covers training AI to support suicide or criminal homicide; provide emotional support or act as a companion; pose as a licensed health professional; imitate human relationships or a human being; or urge someone to isolate or share sensitive information. A violation is a Class A felony. The rule applies to conduct on or after July 1, 2026.
A person harmed by a violation may sue the violator. The right to sue applies to an individual harmed by the prohibited AI training described above. A person under 18, unable to act for themselves, or deceased may have a guardian, estate representative, family member, or court-appointed person act for them. The person may seek actual damages or $150,000 in set damages, punitive damages, and reasonable legal costs. A court may also order the AI to stop operating until the conduct is corrected, and may require new training.
No hearings or votes scheduled.