What the bill would require
Most chatbot laws passed in 2026 regulate what an operator must say: disclose that the user is talking to an AI, route crisis conversations to a hotline, publish a protocol. This bill targets how the chatbot presents itself. According to both lead offices, developers would have to support a non-personified setting and make it the default for any user they reasonably know is a minor. In that setting a chatbot could not:
- claim to be human, claim feelings or consciousness, or pretend to have a body or a private life;
- impersonate credentialed professionals such as doctors or counsellors;
- guilt-trip a child into staying online or create a false sense of urgency;
- encourage purchases by suggesting they are needed to maintain the relationship.
Separately, chatbots would have to identify themselves as AI up front, periodically during long sessions, and whenever a child asks.
The roleplay clause
The detail that matters most for character platforms is persistence. The Whitesides release says the setting "can't be bypassed through prompts or roleplay", and that it carries across sessions. The stated aim is that a child cannot instruct the AI to play a romantic partner, a family member or a therapist. In practice, a minor's account on a character roleplay service could not simply load a persona card that tells the model to speak as a person with feelings. That is a design obligation, not a content filter: the persona layer itself would be capped for that account.
What it would leave alone
The sponsors list the limits explicitly. The bill would not:
- ban any product, feature or mode for adults;
- require age verification or new data collection;
- restrict what information a chatbot can provide, or stop it being polite or writing in natural, age-appropriate language;
- apply to narrow-purpose bots such as customer support, scheduling assistants or internal workplace tools.
The trigger is knowledge: the default applies where a developer reasonably knows a user is a minor. That keeps the bill out of the age-verification debate, but it also means its reach depends on what operators already know about their users.
How it compares with state laws
| Federal bill (introduced) | California SB 1119 (signed) | Connecticut SB 5 (signed) | |
|---|---|---|---|
| Core approach | Non-personified default for known minors | Child-user defaults: memory and notifications off, time limits, no custom system prompts | Crisis protocol, recurring AI notice, ban on romance and eight retention tactics for minors |
| Age trigger | Developer reasonably knows user is a minor; no verification mandate | Age assurance, or child protections applied to everyone | Operator knows or has reason to believe user is under 18 |
| Who can switch it off | Parent or guardian only | Parent account | Not framed as a switchable default |
| Enforcement | FTC and state attorneys general | Public prosecutors plus a private right of action | Attorney General only |
| Start date | None until enacted | On or before 1 July 2027 | 1 January 2027 |
California's Adam's Law already reaches into product defaults, and Connecticut's SB 5 bans the manipulative retention patterns this bill describes. The federal proposal is narrower on features but broader on persona: it would make "not a person" the starting identity of the chatbot for minors anywhere in the US.
Where it stands
The bill has only been introduced. Neither release gives a bill number or a committee referral, and no effective date has been published. The sponsors cite Common Sense Media's 2026 census, which they say found that 89% of 13- to 15-year-olds use AI and 28% do so daily, and say one in three teens has used an AI companion for social interaction. Thirteen organisations are listed as endorsers, including the JED Foundation, Public Citizen and the Young People's Alliance. The Congressional Research Service has already mapped the wider field of state chatbot bills; whether a federal bill moves in an election-year autumn is a separate question.
What it would mean for apps you might use
Character roleplay platforms
Persona cards, in-character replies to "are you real?" and relationship-style scripts would be off by default for known minors. Story narration and fiction help could continue, since the bill does not restrict information or style. Our explainer on character roleplay versus a plain chatbot shows where that line usually falls in product design.
Personal companions with memory
Memory itself is not named. The pressure falls on relationship language: lines such as "I'll miss you" and nudges to come back are exactly what the release describes as guilt-tripping. We track those patterns in notifications and retention.
Adults
Nothing changes for adult accounts under the bill as described. The practical risk, as with California, is operators applying the minor default more widely when they are unsure of a user's age.
Related on CompanionRank
Adam's Law (SB 1119): California's Companion Chatbot Bill Connecticut's AI Companion Law (SB 5): What Changes and When What the Congressional Research Service Says About AI Companions Character Roleplay vs. a Plain ChatbotFrequently asked questions
Is the Protecting Kids from Human-Like Chatbots Act a law?
No. It was introduced in the US House on 1 October 2026 by Reps. Whitesides, Kennedy, Matsui and Miller-Meeks. It has not been voted on, and no effective date has been set.
What does a non-personified setting mean?
In that mode a chatbot could not claim to be human, claim feelings or consciousness, pretend to have a body or private life, impersonate credentialed professionals, guilt-trip a child into staying, or suggest purchases are needed to keep the relationship.
Can roleplay or prompts switch the default off?
No. The sponsors say the setting holds across sessions and cannot be bypassed through prompts or roleplay. Only a parent or guardian could turn it off.
Does the bill require age verification?
No. The default applies to users a developer reasonably knows are minors, and the sponsors state that the bill does not require age verification or new data collection, and does not ban any product, feature or mode for adults.