European Union · Law
EU AI Act, Article 50: chatbots must say they are AI
Applies since 2 August 2026
If people in the EU use a chatbot, it has to tell them it's an AI, at the latest the first time they talk to it, unless that's obvious anyway. The EU pushed back other parts of the AI Act with its 'Digital Omnibus', but not this one.
What it does
- Tell users they are interacting with an AI system
- Provide the notice clearly, no later than the first interaction
- Applies to providers outside the EU too, if the chatbot is used in the EU
- Fines up to €15 million or 3% of worldwide turnover
Sources Jones Walker: August 2 still matters
United States (13 states) · Law
A wave of state chatbot-safety laws in 2026
14 measures passed in 13 states by mid-July 2026; effective dates vary
By mid-2026, states from Washington to Georgia had passed chatbot-safety laws. Most want an 'I'm an AI' notice, a suicide and self-harm protocol, and tighter rules for under-18s, like no sexual content and no tricks to keep them hooked.
What it does
- Washington HB 2225: companion-chatbot safeguards, with a private right of action (signed March 2026)
- Oregon SB 1546: protections for minors from 1 January 2027
- Idaho SB 1297 (from 1 July 2027) and Georgia SB 540 (from 1 January 2027)
- Colorado HB 1263 (from 12 August 2026), Connecticut SB 5 (from 1 October 2026), Hawaii SB 3001 (from 14 July 2026)
- Also Iowa, Nebraska, Rhode Island, South Carolina and Wyoming; New York S 9051 awaited signature
Sources Transparency Coalition: 14 new state laws in 2026
United States · Lawsuits & settlements
Character.AI and Google settle teen-harm lawsuits
Settled (January 2026), terms confidential
Character.AI, its founders and Google settled five lawsuits in Florida, Colorado, New York and Texas. The suits claimed the chatbot played a part in teen suicides and mental-health crises. The terms of the settlement weren't made public.
Sources eWeek: Google and Character.AI settle teen suicide lawsuits · CNN via KVIA (7 Jan 2026)
California, US · Law
California SB 243: companion chatbot safety
In force since 1 January 2026; annual reports from 1 July 2027
If an app works as a companion for people in California, it has to say it isn't human, have a suicide and self-harm protocol, and protect users it knows are minors. If a violation hurts you, you can sue.
What it does
- Clear notice that the chatbot is AI, not human
- Published crisis protocol that refers users to crisis services
- For known minors: a break reminder at least every 3 hours and no sexually explicit material
- Annual reports to California's Office of Suicide Prevention from 1 July 2027
- Private right of action: actual damages or $1,000 per violation
Sources Morrison Foerster: New York and California AI companion laws
Global (company policy) · Company policy
Character.AI ends open-ended chat for under-18s
In effect since late November 2025
After the lawsuits and pressure from regulators, Character.AI shut off open-ended chats with its characters for under-18s and added age checks. Daily limits came first, then the cut-off.
Sources Fortune (29 Oct 2025)
New York, US · Law
New York AI Companion Models Law (GBL Article 47)
In force since 5 November 2025
The first US law written specifically for AI companions. It covers every user, not only minors: the app has to say it's an AI at the start and every three hours, and has to respond to signs of suicidal thoughts by pointing to crisis services.
What it does
- 'You are talking to an AI' notice at the start of each session and every 3 hours
- Detect expressions of suicidal ideation or self-harm and refer to crisis services
- Enforced by the Attorney General: up to $15,000 per day per violation
Sources Morrison Foerster: New York and California AI companion laws · Governor of New York (10 Nov 2025)
United States (federal) · Investigation
FTC inquiry into AI chatbots acting as companions
Opened 11 September 2025
The US Federal Trade Commission ordered seven companies to explain how their companion bots protect kids and teens, how they make money from engagement, and how they test for harm.
What it does
- Orders went to OpenAI, Alphabet, Meta, Instagram, xAI, Snap and Character Technologies
- Answers were due within 45 days
Sources Decrypt: FTC inquiry
Italy · Enforcement
Italy fines Replika's maker €5 million
Decided 10 April 2025
Italy's data protection authority fined Luka Inc., the company behind Replika, €5 million. The reasons: processing user data without a proper legal basis, and age checks it found still weren't good enough.
Sources European Data Protection Board · PYMNTS