Bill > AB1988

This bill requires companion chatbots, which are artificial intelligence systems designed to provide adaptive, humanlike responses and meet a user's social needs, to implement specific safety measures when they detect a "credible crisis expression," meaning a statement that reasonably indicates an intent to harm oneself or others, determined through contextual analysis rather than just keywords. If a chatbot detects such an expression, it must first acknowledge the user's distress, encourage them to seek human support, provide contact information for the 988 Suicide and Crisis Lifeline, and inform them about a potential pause. If the user escalates the crisis or a subsequent crisis expression is detected, the chatbot must initiate a "crisis interruption pause" of 20 minutes, during which it will stop responding, display messages encouraging human support and grounding, and prominently show the 988 Lifeline contact options. The bill also mandates that chatbot operators document all credible crisis expressions and crisis interruption pauses, and starting in 2028, they must annually report this information to the Office of Suicide Prevention. Importantly, chatbots are prohibited from diagnosing users or describing the pause as a punishment.

https://leginfo.legislature.ca.gov/faces/billStatusClient.xhtml?bill_id=202520260AB1988