The laws that matter most right now

The strictest rules apply to companion chatbots, meaning AI that simulates a human relationship. California's SB 243 took effect on 1 January 2026 and, according to law-firm summaries, requires operators to disclose that the user is talking to AI, maintain crisis-response protocols, and add protections for minors, with a private right of action. New York's AI companion law took effect on 5 November 2025 and is reported to require disclosure at the start of an interaction and again at least every three hours, with enforcement by the attorney general.

Other states have narrower or different approaches. Maine requires clear and conspicuous disclosure in certain circumstances, Washington has a companion-bot law reported to take effect on 1 January 2027, and California has had a separate, older bot-disclosure statute for years. Counts of how many states regulate chatbots vary by source, from roughly a dozen to nearly twenty, depending on what is counted.

Several guides note that customer-service bots are excluded from the companion-chatbot definitions. That exclusion describes today's drafts, not a guarantee about tomorrow's bills.

Is your support or sales bot covered?

Usually the companion-chatbot laws are not aimed at a bot that answers order questions or books demos. But there is overlap risk: a branded assistant that chats warmly, remembers users, and is used by minors can drift toward the definitions. And other laws may apply regardless, such as rules against deceiving consumers about whether they are dealing with a human.

The practical test is what the bot is designed to do and how people experience it. If it builds an ongoing, personal rapport, treat it as higher risk and get legal advice. If it answers questions inside a defined task, the disclosure obligation is likely light but still worth meeting.

Disclose at the start

A plain statement that the user is chatting with an AI assistant, shown before the first answer.

Hand off to humans cleanly

Offer a clear route to a person and never claim or imply the bot is one.

A design that satisfies most regimes

Label the assistant as AI in the interface and in the first message. Repeat the disclosure where the law or common sense calls for it, such as in long sessions. Keep a log showing the disclosure was displayed. Add guardrails for sensitive topics and, if minors may use the product, age-appropriate protections.

Build the disclosure as a configurable component so you can change wording or frequency per state without redeploying the whole bot.

Launching a customer-facing assistant? See our chatbot development or talk to us about a compliant design.

Key takeaways

  • California SB 243 (effective 1 January 2026) and New York's companion law (effective 5 November 2025) set the strictest disclosure rules, aimed at companion chatbots.
  • Other states have narrower rules, and counts of covered states vary by source; Washington's reported effective date is 1 January 2027.
  • Customer-service bots are often outside the companion definitions today, but that is not guaranteed to last.
  • Disclose early, offer human handoff, log the disclosure, and make wording configurable by jurisdiction.
This is general information, not legal advice. State chatbot laws are changing quickly and secondary sources differ; check the statutes (for example California SB 243) and current bills before launch.