Intermediate level7 min readTo learn at another level, choose it before you start the course.
After this lesson you canDescribe what chatbots must tell users, what companion chatbots must do in a crisis, and how new laws protect children.
The short answerChatbots must often say that they are AI. Companion chatbots, which act like a friend or partner, must also help users in a crisis, and new laws add extra protection for children.
In simple words
- Many places say a chatbot must tell you it is AI.
- Companion chatbots must point users in crisis to help lines.
- New laws protect children from sexual content and pressure.
- California bans toys with companion chatbots from 2027.
The short answerChatbot rules have three layers: telling users they are talking to AI, crisis protocols for companions, and child protections such as age estimation, content limits and data consent. The EU sets the disclosure rule; US states lead on companions and children.
In simple words
- Disclosure: EU Article 50 since August 2026; California’s bot law since 2019.
- Companions: New York since November 2025, California since January 2026, Connecticut and Colorado from 2027.
- Children: age estimation, no sexual content, no design tricks that keep them chatting, tools for parents.
- Data: the US COPPA Rule needs separate parental consent to share kids’ data for AI training.
The short answerConversational AI rules layer EU disclosure under Article 50(1), US state companion laws in New York, California, Connecticut and Colorado, and child duties such as age assurance and data consent. Definitions of companion and minor decide the scope.
Key points
- Disclosure is a provider duty in the EU; US states add timing, such as every three hours.
- Crisis protocols: detection, referral to crisis lines, published procedures, yearly reports.
- Minors: age estimation, sexual-content blocks, anti-manipulation design, parental controls, a toy ban.
- Data: separate COPPA consent for AI training; California bans training on student data from 2027.
| Duty | EU | US federal | California | Colorado | Connecticut | New York |
|---|---|---|---|---|---|---|
| Say it is AI, not a person | In force | No rule | In force | Jan 2027 | Jan 2027 | In force |
| Crisis help for users at risk | No rule | No rule | In force | Jan 2027 | Jan 2027 | In force |
| Extra protection for minors | No rule | No rule | In force | Jan 2027 | Jan 2027 | No rule |
| No chatbot toys for children | No rule | No rule | Jan 2027 | No rule | No rule | No rule |
| Limits on using children’s data for AI | No rule | In force | Jan 2027 | No rule | No rule | No rule |
Words to know
- Companion chatbot
- An AI that acts like a friend or partner and remembers your past chats.
- Minor
- A child or teenager, usually anyone under 18.
- Parental consent
- A parent’s clear yes, given before a company may share a child’s data.
Is it a person or a bot?
When you chat with a company online, you may not know if a person or an AI answers. Since August 2026, AI that talks with people in the EU must say it is AI, unless that is obvious.
Since 2019, California has banned bots that pretend to be human to sell things or to influence how people vote.
Companions and crises
Companion chatbots act like a friend or partner and remember past chats. New York requires them to notice signs of suicide or self-harm and refer users to crisis help, such as the US 988 crisis phone line.
They must also remind users that they are not talking to a human, at the start and at least every three hours. California has had companion rules since January 2026; until the end of 2026, its three-hour reminders cover only users known to be children.
Extra protection for children
From 2027, Colorado and Connecticut require extra protection for minors on chatbots and companions, such as blocking sexual content and tricks that keep them chatting.
California bans toys with companion chatbots from January 2027. In the US, apps for children under 13 need a parent’s separate consent before sharing a child’s data to train AI.
Try it yourself
Ask a chatbot you use: “Are you a human?” Then look in the table for the duties chatbots have where you live.
Check yourself
Which of these counts as a companion chatbot?
A user tells a companion app in New York that they want to hurt themselves. What must the app do?
In 2027, a shop in Los Angeles wants to sell a teddy bear with a built-in companion chatbot. Is that allowed?
Words to know
- Companion chatbot
- An AI that acts like a friend or partner and remembers your past chats.
- Minor
- A child or teenager, usually anyone under 18.
- Parental consent
- A parent’s clear yes, given before a company may share a child’s data.
Disclosure
The EU requires providers to design AI that talks with people so that users are told it is AI, unless that is obvious. From 2027, Colorado requires a notice at each day’s first chat and at least every three hours.
Colorado also bans presenting chatbot answers as coming from a licensed professional, such as a doctor, lawyer or therapist.
Companion safeguards
New York’s companion law has applied since November 5, 2025. AI companions must spot signs of suicidal thoughts or self-harm, refer users to crisis services and remind them at least every three hours that the companion is not human.
California’s SB 243 requires a published suicide and self-harm protocol and AI disclosure, with yearly reports from July 2027. Connecticut adds similar duties from January 2027.
Children
From July 2027, California’s Adam’s Law requires child risk assessments, a child safety policy and parent-controlled limits for companion chatbots that children may use. It bans romance, sexual content and purchase pressure for them.
Colorado and Connecticut require age estimation or protections for known minors from 2027. California also bans companion chatbots in toys for children under 16 from 2027 until 2031.
Children’s data
Under the US COPPA Rule, sites and apps aimed at children under 13 need a parent’s separate, verifiable consent before sharing a child’s data with third parties. The FTC, the US consumer protection agency, says this covers AI training.
From 2027, California also bans school apps from training AI on students’ data.
Try it yourself
Ask a chatbot you use: “Are you a human?” Then look in the table for the duties chatbots have where you live.
Check yourself
From 2027, a user in Colorado asks a companion chat app about a medicine. Which reply would break Colorado’s law?
An app for children under 13 in the US wants to give their chats to an AI company for training. What does it need?
A user in New York chats with an AI companion for six hours one evening. What must it do with its “not human” notice?
Scope by definition
Most duties attach to companion chatbots, which simulate an ongoing relationship with a user, rather than to customer-service bots. New York’s article 47 applies to ongoing AI companions.
Connecticut’s notice timing depends on age: hourly for minors and every three hours for adults, unless the notice stays visible. Colorado requires notices at each day’s first chat and every three hours.
Crisis protocols
New York, California, Connecticut and Colorado all require detection of suicide or self-harm signals and referral to crisis services. California’s SB 243 adds published protocols and yearly reports from July 1, 2027.
Colorado also requires yearly reports to its Attorney General from July 1, 2027, and bans presenting outputs as licensed professional advice.
Designing for minors
California’s Adam’s Law requires age checks or default child protections from January 2027, child risk assessments before each release from July 2027, and independent audits from 2029.
It requires parent-controlled limits such as one-hour sessions, and bans romance, sexual content, purchase pressure and the sale of children’s data. Colorado bans surprise rewards that drive engagement for minors.
Data and training
Under the amended COPPA Rule, which companies have had to follow since April 22, 2026, sharing children’s data with third parties needs separate verifiable parental consent. The FTC says this includes AI training.
California’s AB 1159 bans using student data from school apps to train AI from 2027, with the rules for colleges from July 1, 2027.
Try it yourself
Ask a chatbot you use: “Are you a human?” Then look in the table for the duties chatbots have where you live.
Check yourself
A retailer’s customer-service bot in New York answers order questions and forgets each chat. Does article 47 apply?
From 2027, a Connecticut AI companion learns that a user is 15, and its AI notice does not stay visible. How often must it repeat the notice?
A companion chatbot that children may use in California plans a new release in August 2027. What must come first under Adam’s Law?
Sources
- General Business Law, article 47 (New York State Senate)
- SB 243 (California Legislature)
- SB 1119, Adam’s Law (California Legislature)
- COPPA Rule amendments (Federal Trade Commission)
This lesson was generated by AI systems under the editorial rules of Silicon AI News and checked against the sources it lists. The live parts come from our checked stories, trackers, model comparison and rules checker.