Sponsors (2)
Requires AI chatbots to disclose AI status, detect high‑stakes or crisis interactions, obtain consent before using conversations for training, report serious incidents to FTC/NIH, and limit recording/retention.
Official title: Establish consumer transparency and protection requirements for artificial intelligence chatbots, and for other purposes.
Introduced July 23, 2026 by Mark Edward Kelly · Last progress July 23, 2026
The bill strengthens transparency, consent, crisis safety, and enforcement for chatbot interactions—especially protecting older adults and people in crisis—but does so by imposing significant compliance, reporting, and enforcement costs and regulatory uncertainty that may slow AI innovation and degrade some user experiences.
All users (including children, young adults, seniors, and people with disabilities) get clearer transparency and stronger control: chatbots must disclose they are AI, obtain affirmative consent before sensitive actions or using conversations for model training, and allow users to delete conversation history.
People in crisis (and the health systems that serve them) are better protected and routed: chatbots must detect crises, refer to emergency services/988, and aggregated incident data and evidence-based practices can improve crisis response and AI safety over time.
Consumers and states gain stronger enforcement and remedy tools: the FTC can use its full rulemaking and penalty powers to enforce key rules while states retain certain enforcement options, and covered entities receive a 60‑day notice-and-cure window for first violations to avoid penalties.
Developers, covered entities, hospitals, and small businesses face substantial new compliance costs, reporting burdens, fines (up to $50,000 per knowing/reckless violation), and ongoing litigation risk that can raise prices or reduce product availability.
AI developers and users may see slower improvements and fewer free/low‑cost features because strict limits on using conversation data for model training, retention, or transcription will constrain data available for improving chatbot quality and personalization.
Delegated, broad, or fast rulemaking by federal agencies (with a one‑year deadline and authority to add “other requirements”) creates regulatory uncertainty and risk of rushed rules, which can be costly to implement and invite legal challenges.
Based on analysis of 9 sections of legislative text.
Requires companies that provide AI chatbots in the United States to disclose when users are talking with an AI, detect and respond differently when conversations involve high‑stakes decisions or crises, and get affirmative consent before using conversations for model training. It also requires annual reporting of serious incidents involving older adults and gives the Federal Trade Commission enforcement authority with civil penalties and state enforcement backstops. Sets standards for accessible crisis responses, limits recording and retention of sensitive conversations, directs NIST to publish voluntary design guidance for chatbots used by older adults, and gives the FTC one year to issue implementing rules and NIH a role in data collection and public reporting.