The States Are Writing the Rules for AI Chatbots

While the frontier labs argue over model safety in the abstract, the concrete rules for how chatbots may talk to people are being written in state capitols. Legislators have introduced close to a hundred chatbot-specific bills across at least 34 states and at the federal level, and more than 140 measures touching chatbots and AI companions moved through the 2025 and 2026 sessions, according to a tally by the Center for Democracy and Technology.

The clearest marker is California's Senate Bill 243, signed in October 2025 and in force since the first of January. It made California the first state to set specific guardrails for companion chatbots, and it has become the template others are copying.

What the California law requires

SB 243 targets the systems designed to hold ongoing, human-like conversations. Operators must tell users clearly that they are talking to software rather than a person. They must maintain protocols for when a user expresses suicidal thoughts, including pointing that person to crisis services. They must shield minors from sexual content and issue regular reminders that a companion chatbot may not be suitable for young users. The law also directs the Federal Trade Commission to publish educational material for parents and teachers, and, notably, it lets families sue developers for harm, as the Future of Privacy Forum has detailed.

That private right of action is what gives the statute teeth. Disclosure rules are easy to write and hard to enforce. Letting affected families take a developer to court changes the calculation for any company deploying a chatbot to consumers in California.

A patchwork, not a policy

The trouble with rules made state by state is that they rarely agree. A company operating nationally now faces a growing set of overlapping and sometimes conflicting obligations, with more arriving each session. Europe has taken the opposite path: the EU AI Act's transparency provisions, which took effect on 2 August, require across the bloc that people be told when they are dealing with an AI system rather than a human, as the European Commission set out.

The urgency behind all of it is the same. As more people turn to chatbots for companionship, for advice, and for support they might once have sought from a professional, the stakes of a bad answer rise. The site has looked at both ends of that shift, from readers asking whether a chatbot can stand in for a therapist to the growing number who ask AI about their health. Lawmakers are trying to draw lines around exactly those uses. For now, where you live decides what protections you get.