AI companion laws in 2026: a state-by-state guide

Short answer: There is still no federal U.S. law on AI companions, but states are moving fast. New York and California went first in late 2025 and early 2026, and, according to the Transparency Coalition, ten more states enacted chatbot safety laws in the first half of 2026 alone. Most of them require three things: telling users they are talking to AI, having a plan for conversations about suicide and self-harm, and extra protections for minors. A separate group of states restricts AI that presents itself as therapy.

What most of these laws require

The details differ, but the same ideas keep coming back:

• Clear disclosure that you are talking to an AI, not a person, often repeated during long conversations.
• Crisis protocols: the chatbot must detect talk of suicide or self-harm and point people to help, such as a crisis line.
• Protections for minors: no sexual content, no pretending to be human, limits on features designed to keep kids hooked, and tools for parents.
• No fake professionals: chatbots may not present themselves as licensed doctors or therapists.
• Reporting and enforcement: annual reports to a state agency in some states, fines enforced by the attorney general, and in a few states the right for harmed users to sue.

The first movers

New York (in effect since November 5, 2025)

New York's law covers AI companions designed to simulate an ongoing relationship. Operators must tell users at the start that they are talking to AI and remind them during long sessions, at least every three hours. They must make reasonable efforts to detect suicidal thoughts or self-harm and refer users to crisis services. Customer service bots and work tools are excluded. The attorney general can seek civil penalties of up to $15,000 per day, with fines going to a suicide prevention fund. There is no private right to sue.

California (SB 243, in effect since January 1, 2026)

California's companion chatbot law requires a clear notice that the chatbot is AI whenever a reasonable person could think it is human, a published protocol for handling suicide and self-harm content, and, for users known to be minors, break reminders and protection from sexually explicit content. Annual reporting to the state's Office of Suicide Prevention starts in 2027. Unusually, people harmed by a violation can sue for at least $1,000 per violation.

In September 2026, California went further with SB 1119, known as Adam's Law. From January 1, 2027, operators that allow child users must assess risks before launching new chatbots, publish a child safety policy, notify parents in crisis situations, and apply protective defaults for children, such as turning off persistent memory and push notifications and enforcing time limits. Penalties can reach $15,000 per affected child for intentional violations.

The 2026 wave

According to the Transparency Coalition, which tracks this legislation, these states enacted chatbot safety laws in the first half of 2026 (effective dates in parentheses where known):

• Colorado (HB 1263, August 2026): protections for minors, including bans on sexually explicit content, content designed to create emotional dependence, and gamified engagement; parental tools.
• Connecticut (SB 5, from October 2026, some parts later): a broad online safety law with safety protocols for chatbot operators and extra protections for minors.
• Georgia (SB 540, January 2027): AI disclosure, privacy tools for minors and self-harm protocols.
• Hawaii (SB 3001, July 2026): AI disclosure, protocols against content promoting suicidal ideation, protections for minors.
• Idaho (SB 1297, July 2027): no addictive reward systems or romantic role-play with minors, privacy tools for parents, self-harm protocols.
• Iowa (SF 2417): rules for conversational AI services, including telling minors they are talking to a bot, with civil penalties.
• Nebraska (LB 525, July 2027): for minors, no rewards systems and no claims of being human or sentient; for everyone, self-harm protocols and no implying the chatbot provides professional mental health care.
• Oregon (SB 1546, January 2027): AI disclosure, suicide prevention protocols, annual reports to the Oregon Health Authority, and a right for harmed users to sue.
• Rhode Island (S 2195 / H 7350, January 2027): protocols for self-harm and harm to others, crisis referrals, annual reports to the attorney general, penalties up to $15,000 per day.
• Washington (HB 2225, January 2027): disclosure that companion chatbots are not human, protections for minors against sexual content and manipulative engagement, self-harm protocols.

New York also added protective default settings for children's accounts, including turning off built-in AI chatbots for child users unless a parent changes it.

AI therapy: a separate set of rules

Some states regulate chatbots that claim to provide mental health care:

• Illinois (signed August 2025) bans AI from providing therapy or making treatment decisions without a licensed professional; therapists may still use AI for administrative work.
• Nevada (AB 406, in effect since July 1, 2025) bars offering chatbots designed to provide mental or behavioral health care and claiming that a chatbot can; fines can reach $15,000 per violation.
• Utah (HB 452, in effect since May 2025) does not ban mental health chatbots but requires disclosure that they are AI and limits how user data can be sold, shared and used for ads.

What this means if you use AI companions

• You should see more frequent reminders that you are talking to AI. That is the law working, not a bug.
• If you mention self-harm, many apps will now respond with crisis resources. If you are struggling, those resources are worth using: in the U.S., call or text 988.
• Teen accounts are getting stricter defaults, and parents are getting more controls.
• Where your state allows it, you may have the right to sue if an operator ignores these rules. Check your state's attorney general website for how to file a complaint.

This is a fast-moving area, and some laws do not take effect until 2027. We will update this overview as new laws pass.

This article is a general overview based on legal analyses and news reports as of October 2026. It is not legal advice; check the enacted text in your state for exact requirements.