AI Companion Chatbot Laws: What Changes January 1, 2027

Three years ago no US state had a statute written specifically for conversational AI companions. By October 2026 there are eighteen of them across sixteen states, and the single busiest date in the whole set is 1 January 2027.

The AI companion chatbot laws now on the books do broadly similar things. They force a bot to admit it is a bot. They require a written plan for what happens when a user starts talking about suicide or self-harm. And they add a thicker layer of protection when the person on the other end is a child. Where the statutes diverge is in the detail, and the detail is where the compliance work actually lives.

What the law means by a “companion chatbot”

The definitions in these AI companion chatbot laws are deliberately narrow. The laws generally carve out customer-service bots, video-game characters with limited scripted dialogue, and voice assistants that answer a question and move on. What they target is software designed to sustain a relationship: a persistent persona, memory that carries across sessions, and conversation that is social rather than transactional.

That boundary matters more than it looks. A general-purpose assistant can slide into scope the moment a product team gives it a name, a personality and long-term memory. Classification is the first compliance decision, not an afterthought.

California: from SB 243 to Adam’s Law

What SB 243 already requires

California’s first companion-chatbot statute, SB 243, was signed on 13 October 2025 and took effect on 1 January 2026. It obliges operators to:

• give a clear and conspicuous notice that the output is artificially generated, where a reasonable person could otherwise be misled into thinking they are talking to a human;
• disclose that companion chatbots may not be suitable for some minors;
• maintain a protocol to keep the bot from producing suicide or self-harm content, refer at-risk users to crisis services, and publish the protocol’s details on the operator’s website;
• file annual reports with California’s Office of Suicide Prevention starting 1 July 2027, covering crisis-referral volumes and detection protocols.

SB 243 also created a private right of action. A person injured by a violation can seek injunctive relief plus the greater of actual damages or $1,000 per violation, with attorney’s fees.

What Adam’s Law adds on 1 January 2027

SB 1119, known as Adam’s Law, was signed on 10 September 2026. It takes effect on 1 January 2027, with most operator duties operative from 1 July 2027, and it replaces SB 243’s known-minor provisions rather than sitting alongside them. According to Hunton’s analysis of the statute, operators must:

• Establish age using specified assurance methods, or apply child protections to every user.
• Run a risk assessment before launch or any substantial modification, and document the mitigations.
• Apply protective defaults to child accounts: persistent conversational memory off (with an exception for users 16 and over), push notifications off, one hour of continuous use and two hours per day. Only a linked parent or guardian can change those settings, and for users under 16 a parent can switch access off entirely.
• Avoid a list of prohibited behaviours. The bot must not claim to be human or sentient, express romantic interest, foster emotional dependence, use excessive flattery, discourage help-seeking or breaks, present itself as a diagnostic or treatment tool outside FDA and HIPAA frameworks, or solicit purchases tied to the relationship.
• Preserve conversations where the operator knows of a credible imminent threat, or that a child died or seriously self-harmed after chatbot conversations: notify a parent where practicable, keep the records for at least three years, and do not delete the account.
• Limit advertising and data use for children, including no cross-context behavioural ads and no sale of children’s chatbot data.
• Submit to independent child-safety audits from 1 January 2029, repeating every two years. Operators below $500 million in prior-year gross revenue are exempt until 1 January 2032.

Enforcement sits with public prosecutors, with civil penalties of up to $5,000 per affected child for negligent violations and up to $15,000 for intentional ones. The Attorney General must stand up a public complaint mechanism by 1 January 2028.

The rest of the AI companion chatbot laws landing on 1 January 2027

California is not alone on that date. A Fenwick survey published on 7 October 2026 counts eighteen enacted chatbot laws across sixteen states. Taking effect on 1 January 2027:

• Connecticut SB 5 — crisis detection and referral, no claiming to be human, AI disclosure, minor protections.
• New York S 9008C (Safe by Design Act) — age assurance before access, AI companions disabled by default for minors with parental override only, parental controls.
• Oregon SB 1546 — AI disclosure, self-harm protocol, minor protections, annual public reporting.
• Rhode Island SB 2195 — detection of self-harm and harm to others with crisis referral, plus annual Attorney General reporting from 1 July 2027.
• Washington HB 2225 — AI disclosure for all users, minor protections, and detection and referral covering suicide, self-harm and eating disorders.
• Colorado HB26-1263 — the statute itself took effect on 12 August 2026, but most operator duties land on 1 January 2027, including age estimation and self-harm referral.

Several laws are already live: Utah HB 452 (7 May 2025), Maine LD 1727 (24 September 2025), New York S 3008C (5 November 2025), California SB 243 and New Hampshire HB 143 (1 January 2026), Tennessee SB 1580 (1 July 2026) and Hawaii SB 3001 (14 July 2026). A second cluster follows on 1 July 2027: Georgia SB 540, Idaho SB 1297, Iowa SF 2417 and Nebraska LB 525.

What AI companion chatbot laws mean if you are a parent

The practical change is that parental controls stop being a courtesy feature. Under Adam’s Law an operator has to provide them, has to let a parent adjust each protective default, and has to actively promote their existence rather than bury them in settings.

Two caveats are worth holding onto. First, the protective defaults only bite when the account is actually identified as a child’s, which is why age assurance is the hinge of the whole scheme. Second, these are product-design and process duties. They oblige a company to detect crisis language and route to help; they are not a guarantee that a model will never produce harmful output.

What this means if you build or sell an AI product

Three jobs come before anything else once AI companion chatbot laws apply to you. Classify the product against each state’s definition, because a chatbot can be in scope in Oregon and out of scope in Maine. Decide the age-assurance posture, since applying child protections universally is a legitimate alternative to estimating ages. Then write the crisis protocol down and publish it, because several statutes require the document to be public, not merely to exist.

The quiet cost is record-keeping. Reports are due to California’s Office of Suicide Prevention and to the Colorado and Rhode Island Attorneys General from 1 July 2027, and to Hawaii’s Department of Health from 1 January 2028. Add preserved conversation records and audit trails, and the reporting calendar becomes its own project. For wider context on the state’s broader rulebook, see our overview of California AI laws in 2026.

What is still unsettled about AI companion chatbot laws

• No federal statute covers companion chatbots, so the patchwork is the regime for now.
• Age-assurance methods are not harmonised across states, and a method that satisfies one legislature may not satisfy another.
• Definitions vary enough to matter. Product teams shipping nationally will likely build to the strictest standard rather than maintain fifty variants.
• Private rights of action in California mean litigation, not only regulator attention, will shape how these duties are read.

If you ship a conversational product to US users, the useful move between now and January is unglamorous: a written classification memo, a documented crisis protocol, and a decision on age assurance you can defend. The deadlines are fixed; the interpretations are not.

AI companion chatbot laws: Key Takeaways

That covers the essentials of AI companion chatbot laws. For more streaming guides and tips, visit KenoIPTV. For further background, see Wikipedia.