Relating to chatbots, including deployer requirements and interactions with minors

This bill relates to public-facing chatbots (public chatbots). The bill defines “AI companion” as a chatbot designed to simulate human-like romantic or emotional bonds. The bill defines “deployer” as a person that makes an AI companion, a public chatbot, or a therapeutic chatbot available to users in this state. The bill defines “public-facing chatbot” as a chatbot intentionally made available to the general public or marketed directly to consumers for independent use without the ongoing supervision of the deployer or an institutional consumer. The bill defines “therapeutic chatbot” as a public chatbot that is designed for the primary purpose of providing mental health support, counseling, or therapy by diagnosing, treating, mitigating, or preventing a mental health condition. The bill also defines “artificial intelligence” (AI), “chatbot”, “commercially reasonable”, “minor”, and “output”. The bill requires a deployer to implement and maintain protocols meant to detect, respond to, report, and mitigate harm the deployer’s public chatbot may cause a user in a manner that takes commercially reasonable steps to protect the well-being of users; limit the collection and storage of user information collected by a public chatbot to what is necessary to fulfill the deployer’s purpose for making the public chatbot publicly available; clearly and conspicuously disclose the public chatbot is AI and not a licensed medical, legal, financial, or mental health professional at the beginning of the public chatbot’s interaction with a user and at three-hour intervals of continuous interaction with the user; and implement protocols to respond to user prompts indicating the user has suicidal ideations or the intent to cause self-harm. The bill prohibits deployers from knowingly or recklessly designing or making a public chatbot available if the public chatbot misleads a reasonable user into believing the public chatbot is a specific human being; misleads a reasonable user into believing the public chatbot is licensed by the state; or encourages, promotes, or coerces a user to commit suicide, perform acts of self-harm, or engage in sexual or physical violence against a human or an animal. The bill requires a deployer of an AI companion or a therapeutic chatbot to implement commercially reasonable measures to determine whether a user is a minor. The measures must use a risk-based approach appropriate with the nature of the public chatbot and the reasonably foreseeable harm that may come from using the public chatbot. Reasonable measures may include self-attestation, technical measures, or other commercially reasonable approaches, but the bill is not to be construed to require a deployer to verify a user’s age using government-issued identification. The bill requires a deployer of an AI companion or a therapeutic chatbot to implement protocols for sending a notification to a minor user’s parent, legal guardian, or legal custodian when the minor user enters a prompt indicating the minor user has suicidal ideations or the intent to cause self-harm. The bill prohibits a deployer from making a therapeutic chatbot available for a minor’s use or purchase unless the deployer meets requirements related to therapeutic chatbot use and safety as detailed in the bill. The bill provides that a deployer is not liable for a user’s misrepresentation of age if the deployer has made commercially reasonable efforts to comply with the bill. The bill authorizes the attorney general to bring actions on behalf of the state to enforce the bill and seek injunctions for violations of the bill. The bill allows a court to issue a civil penalty of not more than $2,500 for each violation of the bill, or $7,500 if the person violated an injunction issued under the bill. Civil penalties assessed under the bill shall be deposited into the general fund of the state. The bill requires the attorney general to notify a person in violation of the bill of the person’s violation and give the person 30 calendar days from the date the attorney general notified the person to cure the violation prior to the attorney general initiating a proceeding to obtain a civil penalty. However, the requirement to notify a person shall not apply if a violation will cause imminent harm to a minor. The bill provides that a deployer who makes commercially reasonable efforts to comply with the bill shall not be subject to liability for unforeseeable or emergent outputs generated by the deployer’s public chatbot. The bill shall be narrowly interpreted to apply only to consumer-facing AI. The bill shall not be construed to regulate internal business technologies. The bill states the legislative intent of the general assembly to support innovation in AI while establishing reasonable consumer protections for systems designed for independent public use.