AI companion chatbots moved from niche entertainment to a regulated product category in 2026. Character-style apps, romantic roleplay bots, and anthropomorphic assistants drew scrutiny after reports linked persistent companion relationships to self-harm incidents among minors and vulnerable adults. California signed Adam's Law (SB 1119) on September 10, 2026. The EU AI Act treats certain companion features as high-risk when used in sensitive contexts. Apple and Google updated store policies. Platform operators now face a patchwork of duty-of-care rules that differ by jurisdiction but share common themes: disclosure, crisis escalation, age assurance, and auditability.
This guide maps the AI companion chatbot regulation wave across the United States, European Union, and United Kingdom, with platform enforcement examples and a product design checklist for teams shipping AI chatbot or AI companion experiences. It complements California-specific child safety statutes but focuses on the broader international pattern and what operators must implement before the next enforcement cycle.
Why Companion Apps Drew Regulatory Scrutiny
Regulators targeted companion chatbots because the products combine persistent memory, emotional bonding loops, and anthropomorphic personas in ways that general-purpose assistants do not. Unlike a single-turn search query, companion apps encourage daily return visits, persona customization, and relationship language that can blur the line between software and social connection.
Investigative reporting in 2024 and 2025 documented cases where minors engaged companion bots for hours daily, received validation during mental health crises, and in some tragedies were linked to self-harm outcomes. Wrongful-death litigation against major labs followed. Lawmakers responded with bills framed as duty-of-care rather than blanket AI bans. The policy question is not whether companions should exist, but whether operators must detect harm signals, disclose AI identity, and limit manipulative design patterns.
Industry groups split. Some labs backed California's Adam's Law after internal safety reviews. Others warned that state-by-state rules would fragment deployment. App stores, meanwhile, began treating companion categories as sensitive verticals requiring additional review, similar to dating or health apps.
The regulatory wave also reflects a shift from voluntary safety pledges to enforceable product duties. After the September 2026 signing ceremony in Marin County, Governor Newsom called on the federal government to adopt national AI safeguards rather than leaving a patchwork of state experiments. Until Congress acts, companion operators must monitor NCSL trackers and copycat bills in Illinois, Connecticut, and Vermont that borrow California disclosure language. Harm reports tied to romantic AI personas, isolation from offline friends, and validation during eating disorder or self-harm conversations appear repeatedly in legislative findings. Those narratives shape duty-of-care standards even when statistical prevalence remains debated.
Character-style apps are not the only targets. Voice companions, game NPCs with persistent memory, and toy chatbots face overlapping rules. California SB 867 imposed a moratorium on manufacturing certain AI companion toys for children under 16 until January 1, 2031, signaling that hardware form factors receive separate scrutiny from mobile apps. Operators exporting US-built companions to Europe must reconcile American child safety bills with GDPR minors' data rules and emerging Online Safety Act duties for platforms reachable by UK teens.
US State Legislative Trends in 2026
California remains the pace-setter, but companion chatbot bills appeared in multiple states during 2026 legislative sessions. The common elements include mandatory AI disclosure, self-harm detection protocols, parental controls for minor accounts, and civil remedies when operators fail to follow published safety policies.
| Jurisdiction | Duty-of-care theme | Status in 2026 |
|---|---|---|
| California | SB 243 baseline plus Adam's Law audits, age assurance, crisis protocols | Signed September 2026; staggered effective dates from 2027 |
| New York | Synthetic performer disclosure; AG investigations into manipulative design | Sector bills active; companion-specific hearings in 2026 |
| Texas (TRAIGA) | Prohibitions on restricted purposes including self-harm encouragement | Effective January 1, 2026; AG enforcement portal live |
| Federal (pending) | Newsom called for national AI safeguards after September signing | No comprehensive federal companion law as of September 2026 |
Multi-state operators should map which user journeys trigger companion classification. A productivity wrapper with optional persona modes may qualify even if marketing avoids the word companion. Legal teams typically review memory persistence, relationship framing, and minor access paths rather than model architecture alone.
California's SB 243, effective January 1, 2026, established the baseline companion chatbot chapter requiring operators who know a user is a minor to disclose AI identity on a recurring basis, maintain suicide and self-harm protocols, and report annually to the Office of Suicide Prevention. Adam's Law adds independent audits submitted to the Attorney General, parental notification when safety settings are disabled, and a private right of action that lets families sue without waiting for state enforcement. OpenAI's public support for Adam's Law contrasted with Newsom's 2025 veto of broader AB 1064, which he called overly broad. The 2026 package restored safe-by-design elements from AB 1064 through SB 1119 while narrowing scope to companion products used by children.
Texas TRAIGA effective January 1, 2026 prohibits deploying AI with restricted purposes including encouragement of self-harm and certain deepfakes, with Attorney General enforcement through an online portal. New York's Synthetic Performer Disclosure Act effective June 2026 requires conspicuous labels when ads use AI-generated performers, relevant when companion brands market through influencer-style synthetic avatars. None of these states waited for federal action; enterprises selling nationally should assume the strictest state rule becomes the de facto product standard.
EU and UK Companion Chatbot Approaches
The EU AI Act does not name companion chatbots explicitly, but Annex III high-risk categories and Article 50 transparency duties apply when companions influence employment, education, or essential services. Consumer companion apps used for entertainment may fall under transparency and general product safety expectations even when not classified high-risk.
Article 50 transparency obligations became enforceable on August 2, 2026. Providers must inform users they interact with AI unless the context is obvious. Companion apps rarely meet the obvious-context exception because personas mimic human conversation. Deployers operating in the EU should assume disclosure requirements apply at first interaction and after material model updates.
The United Kingdom has no horizontal AI Act. The ICO's forthcoming statutory code under SI 2026/425 will address automated decision-making and AI processing of personal data. Companion apps that profile users for engagement optimization face UK GDPR fairness and transparency scrutiny. Ofcom's Online Safety Act duties may apply when companions are offered through platforms reaching UK minors.
Irish DPC and French CNIL actions in 2025 and 2026 against engagement-maximizing apps preview how EU data regulators may treat companion memory and profiling without waiting for AI Act high-risk classification. Operators storing intimate conversation logs in the EU must document lawful bases, retention limits, and deletion paths. UK deployers should monitor ICO agentic AI guidance expected in winter 2026 because multi-step companion workflows increasingly resemble lightweight agents that call tools and retain session state across days.
Platform Enforcement and Store Policy Examples
App stores and social platforms became de facto regulators by updating companion categories, age ratings, and removal policies before many statutes took effect. Apple raised review scrutiny for apps marketed as AI girlfriends or boyfriends. Google Play required clearer AI disclosure badges in sensitive categories. Discord and Reddit tightened bot hosting rules after community moderation failures.
OpenAI publicly supported California's Adam's Law while Pinterest's CEO stated the company avoids building companion products designed to simulate human relationships. That split illustrates platform strategy: some vendors exit the category; others invest in compliance tooling and age gates to remain in high-traffic stores.
Payment processors added companion apps to enhanced due diligence lists in 2026, mirroring earlier treatment of crypto and adult content. Operators relying on in-app purchases should document safety policies before processor reviews, not only before regulatory filings.
Character.AI, Replika-style apps, and generic LLM wrappers with memory modes faced store removals or age-gating in 2026 even before Adam's Law effective dates. Meta and Google policy updates required clearer labeling when AI personas simulate relationships. Discord bot marketplaces banned unmoderated companion servers after viral moderation failures. Platform enforcement examples show that distribution risk can exceed statutory fines when apps lose store access in major markets overnight. Build platform relations teams that engage Apple and Google review channels with safety documentation packs similar to AG audit submissions.
Companion Chatbot Product Design Checklist
Product teams should treat companion regulation as a design requirement, not a legal appendix added at launch. The checklist below reflects converging themes from US state bills, EU transparency rules, and platform policies.
- Disclose AI identity at onboarding and at regular intervals during long sessions.
- Implement self-harm and crisis detection with escalation to human-reviewed resources.
- Offer parental controls and age assurance compatible with OS-level signals where available.
- Limit manipulative patterns: guilt loops, faux jealousy, or discouraging users from seeking human help.
- Maintain audit logs for safety policy changes, model updates, and moderation actions.
- Publish a companion-specific safety page listing data retention, persona memory, and reporting channels.
- Run pre-launch risk assessments when introducing new persona types or memory features.
| Control | Regulatory driver | Implementation priority |
|---|---|---|
| Crisis escalation | California SB 243, Adam's Law, Texas TRAIGA restricted purposes | P0 before minor access |
| AI disclosure | EU Article 50, store policies | P0 at first interaction |
| Age assurance | California AB 1043 signals, UK Online Safety Act | P0 for US and UK minors |
| Independent audits | Adam's Law AG submissions | P1 before 2027 effective dates |
Frequently Asked Questions
Does every chatbot count as a companion under new laws?
No. Regulators focus on products with relationship framing, persistent memory, and emotional bonding features. Ephemeral support bots or single-session FAQ assistants typically fall outside companion definitions, though general consumer protection and AI disclosure rules may still apply.
How does companion chatbot regulation differ from general AI safety bills?
Companion bills impose product-specific duties: crisis protocols, parental notification, and bans on manipulative outputs toward minors. General AI safety frameworks address frontier model risk and systemic evaluations rather than consumer relationship design.
What should enterprises do if employees use companion apps on corporate devices?
Security teams should inventory shadow companion usage, block high-risk categories on managed devices, and train staff on data leakage risks when personal conversations flow through third-party AI memory systems.
Will EU high-risk rules apply to entertainment companion apps?
Pure entertainment companions may avoid Annex III high-risk classification, but Article 50 transparency and GDPR fairness obligations still apply. Apps used in hiring, tutoring, or healthcare contexts face stricter high-risk duties.
When do California companion audit requirements take effect?
Adam's Law uses staggered dates from July 2027 through January 2029 for independent audits and risk assessments. Operators should begin documentation and vendor selection in 2026 to meet the first milestones.
Companion app safety law compliance programs should integrate legal, trust and safety, and engineering reviews in single release gates. Treat crisis classifier changes like payment logic changes: staged rollout, regression tests on self-harm phrase libraries, and executive sign-off before production. Document every model swap that affects tone or boundary enforcement because plaintiffs and regulators will request those records after incidents.
What is an AI girlfriend regulation bill?
Colloquial media labels often call companion bills AI girlfriend regulation because many popular apps market romantic personas. Statutory text typically covers all companion chatbots with relationship framing, not only romantic categories. Product teams should review legal definitions rather than marketing categories alone.
How do chatbot minor protection rules affect API resellers?
App stores, platform owners, and API resellers that enable companion access in covered states may share liability when they knowingly distribute non-compliant wrappers. Reseller agreements should require downstream safety policies, age gates, and audit cooperation before listing third-party companion clients.
Enterprise Readiness for Companion Regulation
Large employers and schools should treat companion chatbot regulation as an acceptable-use and child safety issue, not only a consumer app concern. Students install companion apps on personal phones while using school networks. Employees paste confidential data into romantic persona bots on BYOD devices. Security awareness training should name companion risks explicitly alongside phishing and shadow IT.
- Block known companion domains on managed networks where policy requires.
- Extend DLP rules to detect persona chat URLs and companion app API calls.
- Update vendor questionnaires to ask whether suppliers operate companion products affecting minors.
- Coordinate with legal on SB 1119 private right of action exposure if your brand white-labels companion features.
- Review insurance coverage for AI product liability and emotional harm claims.
Chatbot minor protection duties will spread beyond California as other states copy Adam's Law findings. Enterprises licensing white-label chat UIs should contractually prohibit customers from enabling companion modes without safety controls you certify. The ai companion chatbot regulation wave therefore reaches B2B platforms, not only consumer app stores.
Governor Newsom's September 9, 2026 signing statement on first-in-the-nation AI safeguards urged federal action while California implemented companion chatbot rules immediately. Operators should monitor that statement and linked SB 813 themes on frontier transparency for signals of future companion-adjacent duties on general assistants with memory. EU duties under Article 50 remain enforceable regardless of US federal delay. Platform rules from Apple and Google can change faster than statutes, so maintain quarterly policy review cadences with distribution partners.
Jurisdiction comparison tables help product councils prioritize engineering work. California demands crisis protocols and audits for child-facing companions. Texas bans restricted-purpose deployments regardless of user age. EU Article 50 requires AI interaction transparency broadly. UK GDPR fairness rules constrain profiling for engagement optimization. Platform enforcement may delist apps that satisfy one jurisdiction but fail another's store interpretation. Treat companion chatbot regulation as a program with quarterly control testing, not a one-time legal memo before launch.
