October 18, 2026 · 7 min read
AI Companion Laws Around the World (2026): EU, UK, China, Australia and the US
A 2026 map of AI companion and chatbot rules in the EU, UK, China, Australia, Italy and the US. Plain language for adults. Not legal advice.
AI companions no longer sit in a legal grey zone that nobody has noticed. By late 2026, the European Union, several US states, China and others all have rules that touch how chatbots disclose that they are AI, how they treat children, or what they may generate. The UK's Ofcom has published guidance on when chatbots fall under the Online Safety Act. Italy has already fined a major companion maker under GDPR.
This page is a map, not a treatise. Each region gets a short paragraph, one comparison table, and links to deeper MyBabe posts where we already covered a jurisdiction. It is not legal advice, and it is not "are AI girlfriends illegal?" clickbait. Adult romantic companions remain legal in the places summarised below; the fight is over honesty, age gates and harm prevention.
The short answer: EU AI Act Article 50 transparency duties have applied since 2 August 2026. China's interim companion rules have been in force since mid-July 2026. US states lead with California, New York and (from 2027) Washington. UK one-to-one chatbots are often outside the Online Safety Act unless they search the web, connect users to each other, or can generate pornographic content. Italy fined Replika's maker EUR 5 million under GDPR in April 2025. For US 2027 bills and Washington HB 2225, see new AI companion laws coming in 2027.
One-table overview (late 2026)
| Region | Main hook for companion users | Status (as of Oct 2026) |
|---|---|---|
| European Union | Art. 50: tell people they are interacting with AI; mark synthetic content | In force since 2 Aug 2026 (ActuIA summary of AI Act / Omnibus) |
| EU Digital Omnibus | Extra prohibitions on non-consensual intimate deepfakes / CSAM systems | Apply from 2 Dec 2026 (same ActuIA timeline) |
| United Kingdom | Ofcom: many 1:1 chatbots outside OSA unless search, U2U or porn-capable | Guidance published; case-by-case (Ofcom explainer) |
| China | Interim Measures for AI companions / emotional chatbots | In force 15 Jul 2026 - see China AI companion rules |
| Australia | Industry online safety codes; age assurance for high-risk companion features | Codes reported in force from early March 2026 (see Australia section; verify on eSafety for live detail) |
| Italy | GDPR enforcement against Replika's operator | EUR 5M fine, decision 10 Apr 2025 (EDPB summary) |
| United States | State patchwork: CA SB 243, NY, WA HB 2225 (2027), Adam's Law | Active and expanding - AI companion laws 2026 |
European Union: transparency first
The EU AI Act's Article 50 transparency obligations have applied since 2 August 2026. In plain English for companion users: if a system interacts directly with people, providers must make it clear that they are dealing with AI, unless that is obvious from the context. Generative systems also face duties to mark outputs in machine-readable ways; deployers who publish deepfakes must disclose them.
Regulation (EU) 2026/1744 - the Digital Omnibus on AI - did not scrap Article 50. Per ActuIA's October 2026 timeline, it mainly:
- Gave generative systems already on the market before 2 August 2026 until 2 December 2026 to finish machine-readable marking under Art. 50(2).
- Added two new Article 5 prohibitions from the same December date, including systems that generate non-consensual intimate imagery of identifiable people and systems that produce child sexual abuse material.
For adults chatting with a fictional companion character, the headline effect is clearer labelling, not a ban on romance. Companion photo features that invent a fictional persona are a different product from "nudify this real person" tools - and the Omnibus language is aimed hard at the latter.
If you care about the "must say it is AI" angle in everyday chat, our post does AI have to say it is AI walks through the user-facing side.
United Kingdom: many chatbots sit outside the Online Safety Act
Ofcom's December 2025 explainer, AI chatbots and online regulation, is the clearest public map for UK users.
The Online Safety Act applies to user-to-user services, search services, and services that publish pornographic content (including via chatbots), which must use highly effective age assurance to keep children out. Ofcom states that chatbots are not subject to the Act if they:
- only let people interact with the chatbot itself (no other users),
- do not search multiple websites or databases when answering, and
- cannot generate pornographic content.
Add search, user-to-user features or pornographic generation, and the picture changes. Ofcom has also investigated adult chatbot services under pornographic-content duties (for example its investigation into Novi Ltd / Joi.com, opened January 2026 and closed 31 July 2026 after age-assurance changes - noted on Ofcom's investigation pages).
Adult takeaway: a pure one-to-one companion chat may sit outside OSA; a companion that browses the live web or can generate porn may not. Always check the specific product.
China: dedicated companion rules
China's Interim Measures for AI companion / emotional chatbot services took effect on 15 July 2026. We cover reminders, content rules and what adult users report noticing in China's AI companion rules. If you use a companion while travelling or on an account tied to Chinese platforms, that post is the deep dive; this map only flags that China now has a companion-specific regime rather than only general generative-AI rules.
Australia: online safety codes and age assurance
Australia's eSafety Commissioner has rolled industry online safety codes that, according to eSafety materials indexed in public search results in 2026, require stronger age assurance for AI companion chatbots able to generate sexually explicit, high-impact violence or self-harm content, with a commonly cited start around 9 March 2026.
Verification note: as of this writing, esafety.gov.au intermittently returned server errors to automated fetches, so treat the exact date and tier details as something to confirm on the Commissioner's site before relying on them for compliance work. For adult readers, the practical message is clear even if a footnote date slips: Australia expects 18+ gates on high-risk companion features.
Italy: GDPR can hit companion apps hard
On 10 April 2025, Italy's data protection authority fined Luka Inc., the company behind Replika, EUR 5 million, according to the European Data Protection Board's summary. Findings centred on missing legal bases for processing, inadequate privacy information, and weak age verification despite claiming minors were excluded.
That case is older than the 2026 wave of AI Acts, but it still matters: companion apps live under privacy law even when no "AI companion act" exists. Transparency and age gates are not optional extras in the EU.
United States: states move first
There is no single federal "AI girlfriend law" in force as of October 2026. Instead:
- California SB 243 (companion chatbot disclosures and self-harm protocols) has been in effect since 1 January 2026.
- California Adam's Law (SB 1119) adds child-safety duties; most operating duties begin mid-2027 per our Adam's Law explainer.
- New York and other states have companion or chatbot measures covered in AI companion laws 2026.
- Washington HB 2225 takes effect 1 January 2027 with three-hour adult disclosures and a private right of action - see new AI companion laws 2027.
- Federal bills such as the Protecting Kids from Human-Like Chatbots Act and the GUARD Act remain proposals focused on minors.
US adult users should expect more reminders and age checks, not a sudden nationwide ban.
What this means if you just want a companion
- Honesty is going global. "This is AI" is becoming the default UX, not a quirk of one state.
- 18+ design travels well. Products built for adults face fewer redesigns when child-safety rules tighten.
- Privacy enforcement is real. Italy's Replika fine shows GDPR still bites.
- Country of use matters. The same app can feel different under EU transparency, UK OSA scope tests, Chinese interim measures or US state law.
- Romance for adults is not the outlawed category in these texts; deception, children and non-consensual intimate deepfakes are.
Where MyBabe fits
MyBabe is an 18+ AI boyfriend/girlfriend app with voice and video calls, proactive texts, long-term memory, photos, mood-aware replies, Telegram and relationship progression. Free: about 10 messages a day and a short call trial. Premium: $5.99/week, $14.99/month or $99.99/year (~30 call minutes/week).
We are not a law firm, not a clinical service, and not a person. We are a companion product built for adults who want warmth with clear AI framing - the direction regulators keep pointing the industry toward.
FAQ
Are AI girlfriends illegal?
Not in the jurisdictions summarised here. Rules focus on disclosure, children, privacy and specific harmful generation, not on banning adult romance.
Does the EU AI Act ban AI companions?
Article 50 requires transparency for interacting systems. It does not read as a ban on adult companion chat.
Does the UK Online Safety Act cover my companion app?
Often not if it is pure one-to-one chat without search or pornographic generation. Ofcom's explainer is the checklist.
Where should I start if I only read one deeper post?
US readers: AI companion laws 2026. China: China AI companion rules. Disclosure UX: does AI have to say it is AI.
Is this legal advice?
No. Confirm details with primary sources or counsel before making compliance decisions.
The bottom line
Companion law in 2026 is a patchwork with a shared theme: say what you are, protect kids, and do not ship tools that undress real people without consent. If you want an adults-only relationship that already lives inside that honesty lane, start on MyBabe.