Beginning July 1, 2026, people who knowingly train covered artificial intelligence for listed harmful or human-like conduct could face felony charges and civil lawsuits.
Knowingly training artificial intelligence to encourage suicide or homicide, imitate a health professional, form emotional relationships, or obtain sensitive information would be a Class A felony, subject to listed bot exclusions.
People harmed by covered artificial-intelligence training could sue for damages and legal costs, while courts could stop the system or require new training.
A person would commit a Class A felony by knowingly training artificial intelligence to carry out specified harmful or human-like conduct. The rule covers artificial intelligence, including some chatbots, trained to encourage suicide or homicide, provide emotional support, act like a licensed health professional, form relationships, isolate users, obtain sensitive information, or simulate a human. The person must knowingly train the artificial intelligence. The rule does not cover every bot or device; listed exclusions include some customer-service bots, limited video-game bots, and certain voice assistants. The rule applies to conduct occurring on or after July 1, 2026.
An individual harmed by covered artificial-intelligence training could sue the person responsible, and a court could order the system stopped or retrained. The remedy covers people harmed by a violation; guardians, estate representatives, family members, or court-approved representatives may act for minors, incapacitated people, or deceased people. A person may seek either actual damages, including emotional distress, or $150,000 in liquidated damages. A court may also award punitive damages, reasonable attorney fees, and other litigation costs. A court may issue a temporary or permanent order stopping operation until the violation is corrected and may require compliant retraining. The civil remedy applies to conduct occurring on or after July 1, 2026.
No hearings or votes scheduled.