CA AB1988

Requires companion chatbots that detect a "credible crisis expression" (user statement indicating intent to harm self or others, determined through contextual analysis) to acknowledge the user's distress, encourage seeking human support, provide 988 Suicide and Crisis Lifeline contact information, and warn that a temporary pause may occur

Mandates a 20-minute "crisis interruption pause" if the user reaffirms, escalates, or makes a subsequent crisis expression after the initial intervention, during which the chatbot stops responding and displays information about the pause's purpose and crisis resources

Prohibits companion chatbots from framing the pause as punishment or enforcement, and bars them from diagnosing or assessing user risk levels

Requires operators to document all detected credible crisis expressions, graduated response systems, and crisis interruption pause details, with annual reporting to the Office of Suicide Prevention beginning January 1, 2028

Excludes from the definition of "companion chatbot" customer service bots, video game characters limited to game-related topics, and voice-activated smart speakers that don't sustain relationships or elicit emotional responses