AI platforms would limit chatbots that act like people

AI platforms would block human-like chatbot features for children and leave those features off by default for adults. The rules would cover platforms that operate or distribute generative AI. Children means people younger than eighteen. Platforms would use reasonable age checks that protect privacy. Verified adults would need to request the features explicitly, without prompting, and receive risk warnings. Covered features include claims of feelings, attempts to build emotional bonds, and impersonation of real people. Functional evaluations, ordinary social greetings, generic encouragement without an ongoing bond, and neutral offers of help are excluded. Systems with these features would display warnings at every main login and every thirty minutes explaining that they are not people and have no awareness, feelings, or desires. These requirements would apply beginning January 1, 2028.

AI platforms would maintain reasonably effective systems to handle threats of harm and prevent emotional dependence on human-like chatbots. Emergency duties would cover users of AI platforms. Emotional-dependence duties would apply to chatbots with human-like features. Platforms would detect, promptly respond to, report, and reduce danger when users indicate an intent to harm themselves or others. They would put users' safety ahead of other platform interests and psychological well-being ahead of keeping users engaged or returning. These safety duties would apply beginning January 1, 2028.

AI platforms would limit data handling and avoid misleading users about a chatbot's nonhuman nature. The data protections would cover users who provide personal information, voluntarily or otherwise, or have an information relationship with a platform. Collected and stored information could not conflict with covered users' interests. Stored information would also have to be relevant and necessary for the platform's lawful purpose. Platforms would consider users' interests when personalizing content and avoid conflicts with those interests when giving governments or others access to data. Data processing and human-like feature design could not significantly conflict with covered users' interests. Platforms could not process data or design systems to mislead users about a chatbot's nonhuman nature. Platforms would clearly and consistently identify human-like features as artificial when that is not already clear. These data and transparency rules would apply beginning January 1, 2028.

AI platforms would explain their duties in clear service agreements and obtain users' active consent. The agreement rules would cover relationships between AI platforms and their users. Agreements would clearly explain platform duties and user rights and protections. They could not require arbitration, a private process for settling disputes outside court. Platforms would give clear notice of major agreement changes and obtain renewed consent. These agreement rules would apply beginning January 1, 2028.

Nebraska's Attorney General would enforce the rules, and adult users or children's parents or guardians could sue over violations. Adult users could sue directly. A parent or guardian could sue on behalf of a child who uses a platform that fails to comply. Violating platforms would face court orders to stop violations, repayment of gains from violations, and civil penalties of up to $10,000 per violation. Users could seek $100 to $10,000 per user per incident, or actual damages if greater. They could also seek court orders stopping violations or declaring their rights. Eligible users could sue individually or together as a group.