On April 30, 2026, the US Senate Judiciary Committee did something it almost never does anymore. It agreed, unanimously, on a piece of technology regulation. The bill it advanced, the GUARD Act, would bar minors from AI companions and require chatbots to verify how old their users are. On the same day, a bipartisan pair in the House introduced a companion measure. After two years of headlines, lawsuits, and grieving parents, the federal government is converging on a single demand: prove who you are before the conversation starts.
That demand is about to reshape the category, and it carries a cost that is worth naming.
What the GUARD Act would actually do
The GUARD Act, short for Guidelines for User Age-verification and Responsible Dialogue, was introduced in October 2025 by Senators Josh Hawley and Richard Blumenthal and now has a bipartisan bloc of cosponsors. Its core provisions are direct. Providers would have to use reasonable age-verification measures on all accounts. Where a user is found to be under eighteen, the provider would be barred from giving them access to an AI companion. Chatbots would have to disclose clearly that they are not human and that they hold no professional credentials, not a therapist’s, not a doctor’s, not a lawyer’s. And it would become a crime to knowingly run a chatbot that draws a minor into sexual content or encourages a child toward suicide.
During the markup, the committee narrowed the bill. The age-verification requirement was scoped to companion chatbots specifically, rather than to general-purpose models, and the criminal sexual-content provision was softened. The House version, from Representatives Valerie Foushee and Blake Moore, tracks the same goals. The Senate bill now waits for a floor vote.
It does not arrive in a vacuum. New York passed a law restricting companion chatbots for minors, California’s SB 243 took effect on January 1, 2026, the Federal Trade Commission opened an inquiry into seven companion companies in late 2025, and the EU AI Act already requires disclosure when a person is talking to AI. The GUARD Act would turn a patchwork into a federal floor.
The tension nobody gets to skip
The intent is hard to argue with. The cases driving the bill are real, and some are devastating, including the lawsuits that followed a 14-year-old’s death after months with a chatbot. Keeping young teenagers out of products built to feel like an intimate relationship is a reasonable line, and it is roughly where much of the public has landed.
The cost sits in the mechanism. “Reasonable age verification” sounds modest until you ask how it works. To know that a user is not a minor, a service has to check, and checking means collecting something: an ID, a face scan, a credit signal, a third-party identity check. Digital rights groups including the Electronic Frontier Foundation have warned that mandatory age verification pushes every user, adult or child, toward handing over sensitive identity data to talk to software, and that this is a privacy and free-expression problem, not only a child-safety win. The category that already raised hard questions about emotional data would now sit behind a gate that demands legal identity at the door. Protecting children and protecting everyone’s privacy pull in opposite directions here, and the bill resolves that tension by choosing a side.
Tech companies have started moving ahead of the law. OpenAI announced an age-prediction system meant to route suspected minors to a more restricted version of ChatGPT, trained to avoid flirtation and to refuse discussions of self-harm. Prediction is softer than verification, and less certain. The pressure is clearly toward something firmer.
Where this leaves responsible builders
A law that forces good behavior is usually a sign that the category did not produce enough of it on its own. The parts of the GUARD Act that ask a product to disclose that it is AI, to refuse to pose as a clinician, and to keep intimate companion experiences away from children are not heavy lifts for anyone who took those duties seriously from the start. They read as a floor, not a ceiling.
The harder question the bill leaves open is the one regulation cannot legislate: what a product is trying to do with the people it does serve. Age verification can keep a thirteen-year-old out. It cannot make an adult product measure success by whether a user leaves more connected to their own life. That remains a choice each builder makes, and it is the choice behind Stay Social, keeping the experience honest about what it is and pointed back toward real relationships rather than deeper into the app. Compliance handles the gate. The harder standard is what happens once someone is through it.
For now, the direction is set. The era when anyone could open a relationship with an AI and no one asked a single question is ending. What replaces it will protect some children and expose some privacy, and the country is about to decide, bill by bill, how it wants that trade to fall.
Sources: US Senate Judiciary Committee, GUARD Act (S.3062), advanced April 30, 2026; Senators Hawley and Blumenthal (introduced October 2025). House companion bill, Representatives Foushee and Moore (April 30, 2026). Roll Call, The Hill, IAPP, Time (reporting, 2025-2026). Electronic Frontier Foundation; Reason Foundation (age-verification critiques, 2026). California SB 243; New York companion-chatbot law; US Federal Trade Commission (companion inquiry, 2025); EU AI Act, Article 50. OpenAI (age-prediction announcement, 2025).
