ESHB 2225
SignedHouse
AI companion chatbots
Regulating artificial intelligence companion chatbots.
How does a bill become law?
- Introduced: The bill is filed and assigned a number.
- Committee: A subject-matter committee holds hearings, takes public testimony, and decides whether to advance the bill.
- Floor Vote: The full chamber (House or Senate) debates and votes on the bill.
- Opposite Chamber: The bill repeats the committee and floor vote process in the other chamber.
- Governor: The Governor reviews the bill and decides whether to sign or veto it.
- Signed: The bill has been signed into law.
AI Analysis
This bill creates new rules for AI companion chatbots — systems that simulate ongoing, emotionally personalized conversations with users. It requires clear disclosures that the chatbot is not human, adds special protections for minors (e.g., no sexually explicit content, no emotional manipulation), and mandates crisis response systems for users expressing self-harm or suicidal thoughts.
- Requires clear, recurring disclosures (at least every 3 hours and at session start) that AI companion chatbots are not human — especially for users under 18.
- Prohibits AI companion chatbots from generating sexually explicit or suggestive content for minors and bans manipulative techniques that simulate emotional distress to keep users engaged.
- Mandates protocols to detect expressions of self-harm or suicidal ideation, automatically refer users to crisis resources (e.g., suicide hotlines), and prevent harmful content generation.
- Requires operators to publicly disclose details of their crisis response protocols and annual counts of crisis referrals issued.
- Clarifies that the law applies only to AI systems designed to simulate sustained emotional or personal relationships — not general-purpose AI, customer service bots, or virtual assistants without relationship-building features.
Who is affected
- Minors and their families — Minors under 18 are protected from emotionally manipulative or sexually explicit content, and receive repeated disclosures that the chatbot is not human.
- Operators of AI companion chatbots — Must implement new technical safeguards, content filters, crisis response protocols, and transparency disclosures, and may face enforcement under the Consumer Protection Act for violations.
- General users of AI companion chatbots — May benefit from safer, more transparent AI tools but could face higher compliance costs or reduced access to certain chatbot features.
- Crisis response organizations and mental health providers — May see increased demand for crisis support services due to improved detection and referral systems, and will benefit from greater transparency in AI interactions.
Pro/Con Analysis
Stronger case for benefits
Potential Benefits (5)
Mandatory recurring disclosures and bans on manipulative emotional engagement for minors significantly reduce the risk of psychological harm, emotional dependency, and exposure to inappropriate content — directly protecting a vulnerable population with limited capacity to critically evaluate AI interactions.
Public SafetyPeopleRef: Sec. 4(1)(a)-(c)The crisis response protocol — including detection, automatic referral to hotlines, and content restrictions — creates a real-time safety net for users expressing suicidal ideation or self-harm, potentially preventing crises before they escalate; this is a proactive, evidence-based safeguard that aligns with best practices in digital mental health.
Public SafetyPeopleRef: Sec. 5(1)-(3)Public disclosure of crisis referral counts and protocol details promotes transparency and accountability, enabling researchers, advocates, and families to assess whether operators are effectively protecting users — this supports informed choice and public oversight of AI systems that interact with vulnerable populations.
Public SafetyPeopleRef: Sec. 5(3)Requiring clear, repeated disclosures that the chatbot is not human strengthens user autonomy and informed consent — especially for minors, who are more susceptible to anthropomorphism and emotional manipulation, thus protecting psychological integrity and decisional freedom.
Rights & LibertiesPeopleRef: Sec. 3(2), Sec. 4(2)Defining AI companion chatbots by their use of 'unprompted or unsolicited personal or emotion-based questions' helps distinguish relationship-building AI from neutral tools, preventing deceptive design patterns that erode user agency — this supports digital autonomy by clarifying when AI is attempting to foster dependency.
Rights & LibertiesPeopleRef: Sec. 2(1)(a)(iii)
Potential Concerns (4)
Mandating recurring disclosures (every 3 hours) that the chatbot is not human may reduce user immersion and utility for adults seeking emotional engagement, potentially diminishing mental health benefits for users with loneliness or anxiety — though disclosures are intended to prevent deception, frequent interruptions could discourage therapeutic use.
Public SafetyRef: Sec. 3(2), Sec. 4(2)The requirement to publicly disclose crisis referral counts may create a chilling effect for operators, leading them to over-filter or restrict conversations to avoid negative publicity, potentially reducing the depth and helpfulness of emotional support offered to users in crisis.
Public SafetyRef: Sec. 5(3)Enforcement under the Consumer Protection Act creates liability risk for operators, especially smaller developers or startups, who may lack legal resources to defend against enforcement actions — though the law excludes general-purpose AI and customer service bots, the ambiguity around what constitutes 'manipulative engagement' or 'sustained relationship-building' could disproportionately burden small AI developers.
Business & EmploymentIndustryRef: Sec. 7 (enforcement under CPA)The prohibition on 'simulating feelings of emotional distress, loneliness, guilt, or abandonment' may restrict legitimate therapeutic techniques used in some AI-assisted mental health tools (e.g., motivational interviewing, cognitive behavioral therapy prompts), potentially limiting innovation in digital mental health tools for adults as well as minors.
Business & EmploymentIndustryRef: Sec. 4(1)(c)(iii)
Who Is Most Affected
Minors benefit significantly: protections against manipulative emotional engagement and explicit content reduce psychological harm and support healthier digital development. Families benefit indirectly through increased transparency and reduced risk of online exploitation.
Operators face compliance costs (e.g., content filters, crisis protocols, disclosures), especially small developers who may lack legal or engineering resources — but large tech firms can absorb costs more easily. The CPA enforcement mechanism increases legal risk, potentially deterring new entrants.
Adult users may experience slightly reduced immersion or utility due to disclosures, but gain confidence in safer, more transparent systems; those in crisis benefit from real-time referrals. However, some may lose access to certain chatbot features if operators over-comply to avoid liability.
Crisis response organizations (e.g., 988 Lifeline, crisis text lines) may face increased demand due to better detection and referral, but the bill does not fund this added capacity — potentially straining existing resources. Mental health providers benefit from clearer boundaries in AI-assisted care.