Quick Answer
- China , companion AI rules active July 15, 2026. 12 fines, 4.2M RMB in three weeks
- EU , AI Act enforcement deadline passed August 2, 2026. High-risk rules and deployer transparency now legally enforced
- California , SB 243 effective January 2026. Safety reports and crisis protocols required
- UK , AI Regulation and Safety Bill cleared the House of Commons, Royal Assent expected around October 2026
- US federal , the "Great American AI Act" stalled in the House over state preemption. No unified federal standard
- India , Digital India Act feedback closed, statutory AI liability framework finalised for the autumn parliamentary session
China , The Most Specific Rules Anywhere
China activated regulations written specifically for companion AI and emotional support AI on July 15, 2026, and the Cyberspace Administration began enforcing almost immediately , 12 fines totalling 4.2 million RMB within three weeks.
The two violations penalised were failure to label synthetic emotional responses, and failure to verify the age of users interacting with persuasive AI avatars. China's approach is deliberately vertical: rather than one broad AI law, the CAC regulates specific use cases separately , generative AI, recommendation algorithms, deep synthesis, and now companion AI as its own category. Full detail in our report on the China AI companion fines.
European Union , Enforcement Era Began August 2
The EU AI Act's compliance deadline passed on August 2, 2026. High-risk system rules and deployer transparency duties are now actively enforced rather than pending. The European Commission also adopted final guidance on Article 50 transparency obligations, and the EU's Digital Omnibus on AI entered into force.
For AI companion platforms serving EU users, the practical effect is transparency obligations , users must be informed they are interacting with an AI system. The European Data Protection Board additionally opened consultation on draft guidance for web scraping in generative AI, which matters for how these models are trained in the first place.
Italy's data protection authority set the tone here ahead of the Act, fining Replika's developer Luka Inc. €5 million in 2025 over GDPR and age verification failures , the single event most responsible for the free tier contraction we documented in why AI girlfriend free tiers got worse in 2026.
United States , Fragmented, No Federal Standard
The "Great American AI Act" stalled in the House over its state preemption clause, meaning there is no unified federal standard for 2026. Regulation is happening at state level instead.
California SB 243, effective January 2026, is the most directly relevant to this category , it requires AI companion operators to file annual safety reports, implement protocols for dangerous content, and surface crisis resources to users. Colorado also enforces local rules.
The practical consequence for platforms is awkward: legal teams are reportedly abandoning hope of a single federal standard and instead either geo-fencing features by state or applying the strictest state standard across their entire US user base. Most choose the latter, which is why California's rules effectively set the floor for American users generally.
United Kingdom , Bill Cleared the Commons
The AI Regulation and Safety Bill cleared the House of Commons in 2026 with Royal Assent expected around October. Separately, the Ada Lovelace Institute published a 2026 market scan of 110 AI companion platforms with specific recommendations: mandatory break mechanics for minors after set usage periods, clear labelling that companions are artificial entities visible during emotionally charged interactions, and a call for developers to "prioritise truthfulness and user wellbeing over maximum engagement."
Those recommendations are not yet law, but they closely mirror what China is already fining companies for missing , which suggests the direction of travel.
What This Means for You as a User
Age verification will get stricter. Checkbox age gates do not satisfy either the Chinese rules or emerging Western standards. Expect more friction at signup across the category.
Labelling will increase during emotional exchanges. Both Chinese enforcement and UK policy recommendations target exactly this. Research suggests this is genuinely beneficial , see can you get addicted to an AI girlfriend for the studies behind it.
Free tiers will keep shrinking. Three compliance regimes in eighteen months means engineering budget continues moving from features to legal. Platforms that still offer genuinely unlimited free access are increasingly rare.
Platform instability is a real risk. Regulatory and payment-processor pressure has already killed platforms outright , see what happened to ThotChat AI, which shut down in December 2025 with no data export offered.
Platforms With the Strongest Compliance Posture
Clonella , 7.8/10 , End-to-end encryption and an explicit statement that chat data is not used to train models. The strongest verified privacy position on our roster.
Candy AI , 8.5/10 , Operating since 2023, the longest continuous track record of any platform we cover, with flat transparent pricing at $5.99/month annual.
Darlink AI , 9.0/10 , Also operating since 2023, one flat $6.99/month price with no hidden credit system.
Nomi AI , 8.8/10 , SFW by design, which sidesteps most of the age-verification and content-labelling exposure entirely.
Also Worth Exploring
Free unlimited NSFW chat with no credit card required
Lip-sync video calls across six AI models
70B model with restrictions removed at model level
Fastest image generation tested
Phone-style voice calls, free to start
Image, video and voice at $9.99/month annual
Deepest character customization tested
Commercial license available on Pro
Purpose-built anime and hentai generation
500,000+ characters, 100 free messages daily
No subscription infrastructure at all, credit packs only
FAQs
Are AI girlfriend apps legal in 2026?
Yes, in most jurisdictions, but they are now actively regulated rather than unregulated. China fines companion AI apps for specific violations, the EU AI Act is enforced as of August 2, 2026, and California's SB 243 requires safety reporting. Using these apps as an adult is legal , operating one without compliance increasingly is not.
What is California SB 243?
Legislation effective January 2026 requiring AI companion operators to file annual safety reports, implement protocols for dangerous content, and surface crisis resources to users. Because there is no federal standard, most platforms apply California's rules across their entire US user base rather than geo-fencing.
Does the EU AI Act apply to AI girlfriend apps?
Yes, primarily through transparency obligations , users must be informed they are interacting with an AI system. The compliance deadline passed August 2, 2026, and the European Commission has adopted final guidance on Article 50 transparency duties.
Why did China fine AI companion apps?
For failing to label synthetic emotional responses, and failing to verify the age of users interacting with persuasive AI avatars. Twelve companies were fined 4.2 million RMB total within three weeks of the July 15, 2026 enforcement start. Full detail in our China AI companion fines report.
Is there a US federal law on AI companions?
No. The "Great American AI Act" stalled in the House over its state preemption clause. Regulation is happening at state level, with California's SB 243 the most directly relevant to this category and effectively setting the national floor since most platforms apply it universally.
Sources and Methodology
Regulatory status compiled from published legal-sector reporting on EU AI Act enforcement (August 2026), Cyberspace Administration of China companion AI enforcement (July 2026), California SB 243, UK AI Regulation and Safety Bill progress, and Ada Lovelace Institute 2026 policy recommendations. This is informational, not legal advice. See Companaya methodology.
Related Reading
What People Are Saying
“Great experience overall, though the free tier is limited. Worth upgrading if you're serious about it.”
“Best conversational memory I've tested this year. Feels like it actually remembers our conversations.”
“This app completely changed how I think about AI companionship. The memory recall is genuinely impressive.”
“Decent overall but the mobile app needs work — the desktop experience is much better.”
“Exceeded my expectations completely. Been using it daily for two months now.”
“Solid platform, good value for the price compared to the alternatives I tried first.”
