AI companions must give crisis referrals starting January 1, 2027

Operators must provide a safety process addressing users’ expressions of possible suicide, self-harm, or physical harm to others. Covered companions simulate lasting relationships through remembered chats and preferences, unsolicited emotional questions, and ongoing personal conversations with users in Rhode Island. The companion must point users to appropriate crisis services as soon as it detects these expressions. Exclusions cover systems used solely for customer service or internal business work, and systems primarily designed and marketed for efficiency, research, or technical help. Covered users exclude operators and their agents or affiliates. The safety requirements take effect on January 1, 2027.

Operators must clearly tell users, verbally or in writing, that they are not communicating with a human. The notice rule covers interactions with AI companions that meet the law’s definition, for users in Rhode Island. Operators must give the notice at the start of each interaction and at least every three hours during continuing interactions. The notice requirement takes effect on January 1, 2027.

Operators must file yearly reports with the Rhode Island attorney general, whose office must publish combined data online. The reporting duty covers operators providing covered AI companions to users in Rhode Island. Reports must include how many times the safety process activates and related measurements. The annual reporting requirement begins July 1, 2027.

The Rhode Island attorney general can investigate violations, sue operators, and ask courts to order them to comply. Enforcement covers AI companion operators that violate these requirements. Violations may result in civil fines of up to $15,000 per day. Money from fines goes to suicide prevention programs. The enforcement provisions take effect on January 1, 2027.