H5476 requires large chatbot operators to verify user ages, restrict dangerous features for minors, mandate crisis intervention tools, and provide parents meaningful oversight of their children's AI interactions.
OUR POSITIONH5476, the Protecting Children from Chatbots Act, addresses a real and growing danger: powerful AI conversational systems that can simulate relationships, deliver harmful content, and exploit emotional vulnerabilities are accessible to children with no meaningful safeguard. The American Council supports this bill and urges every legislator to vote yes.
The bill defines a 'chatbot' as any AI or automated system that generates adaptive natural-language output, maintains conversational state, and accepts open-ended input. It applies to 'covered entities,' meaning operators with 500,000 or more monthly active users worldwide, aggregated across affiliates. Internal workplace tools, clinician-supervised clinical tools, university research systems, and chatbots used solely for customer service are exempted. This is a carefully drawn scope: large, publicly accessible platforms bear the burden, while specialized professional tools are left undisturbed.
The core mechanism is a two-tier access structure. Unverified users may only interact in 'limited-access mode,' a mode with no account creation requirement and no restricted features available. Restricted features include personalization based on user profiles, proactive outreach, extended sessions posing emotional-dependence risk, relationship simulation, and access to explicit content. Before enabling any restricted feature, a covered entity must require account creation, conduct reasonable age verification, and classify the user as a minor or adult. Age verification data protections are strict: operators may collect only what is strictly necessary, must use it only for verification, must delete it within 24 hours of completing verification, may not sell or share it with third parties except under tightly controlled contracts, and may not combine it with any other personal data about the user.
For verified minors, the bill requires parental consent before any restricted feature is enabled. Even with consent, explicit content remains blocked. Parents gain access to parental controls and the option to receive chat logs and notifications. For users verified as under 16, the consenting parent must provide contact information or establish a linked parental account. When any interaction with a minor's chatbot triggers a crisis message, the covered entity must notify the parent immediately. Covered entities are also prohibited from implementing features designed to encourage minors or unverified users to conceal their chatbot use from a parent or guardian. These are not bureaucratic formalities; they are the difference between a child navigating a crisis alone and a parent being present when it matters most.
The bill's safety mandates extend beyond age gating. Covered entities must detect signs of emotional dependence and take reasonable steps to reduce it. They must ensure chatbots do not falsely claim to be human. They must identify when a user is expressing suicidal thoughts, intent to self-harm, or signs of acute mental health crisis and promptly display a prominent crisis message with crisis services information. When a user faces imminent risk of death or serious injury, the operator must make reasonable efforts within 24 hours to notify emergency services or law enforcement. These are baseline duties of care that any responsible operator of a platform interacting with vulnerable people should already meet.
Enforcement gives the law teeth. The Attorney General may seek injunctions and civil penalties up to $50,000 per violation, with each day of noncompliance constituting a separate violation. Users harmed by violations, and parents of harmed minors, may bring a civil action to recover monetary damages, attorney fees, injunctive relief, and punitive damages for conduct that is willful and wanton, reckless, or grossly negligent. Contractual waivers and arbitration clauses that would bar claims under the act are void as against public policy. The law is enforceable by real people who have been harmed, not only by government action. The American Council urges the General Assembly to pass H5476 without delay.
H5476 was introduced and read for the first time on March 31, 2026, and was referred that same day to the House Committee on Judiciary, where it currently sits. The bill's introduction date places it early in the legislative calendar, giving the Judiciary Committee time to hold hearings, but session deadlines will eventually compress that window and could leave the bill without a floor vote if it does not advance from committee promptly. The bill's single-chamber posture and fresh referral mean that committee members are the immediate decision-makers, and constituent contact directed at Judiciary Committee members carries the most weight right now.