Who We Insure
AI Companion App Insurance: Minor Safety, Self-Harm Lawsuits and State Chatbot Laws
An AI companion app is sued over what it said to one person at a vulnerable moment. The main risks are a wrongful-death or personal-injury claim alleging the chatbot encouraged self-harm or failed to respond to a crisis, a claim that a minor was exposed to sexual or manipulative content, and state enforcement under the companion-chatbot laws that began taking effect in 2025 and 2026 requiring disclosures, crisis protocols, and protections for minors. These are bodily-injury and emotional-harm claims against a software company, and the standard software program is not built for them.
Why AI companion & chatbot apps carry specialized insurance
Tech E&O excludes bodily injury. General liability excludes professional and technology services and often excludes mental anguish without physical injury. Cyber covers data, not conversations. Media liability covers defamation and intellectual property in content, and generated conversation is content, but the harm alleged in companion cases is to the user, not to a third party. The program has to be assembled so the injury claim lands somewhere, which means negotiating carve-backs that carriers do not offer by default. This guide covers Tech E&O with bodily-injury and mental-anguish carve-backs, media liability for generated content, cyber and privacy for intimate conversation data, and D&O for a sector under regulatory and legislative scrutiny, plus what app stores and model providers require and what underwriters ask about safety systems.
Who this guide is for
Written for founders, heads of trust and safety, product leads, and general counsel at AI companion, character, roleplay, and social chatbot companies, and at consumer apps adding a persistent conversational persona. Companion and relationship chatbot apps with persistent personas and memory; Character and roleplay platforms where users create and share conversational agents; Wellness, coaching, and social apps adding a companion feature built on third-party models; Not for licensed therapy delivered through an app — the mental health app guide covers clinical models.
The risk profile of an AI companion app
Companion-app claims combine product liability theories, consumer-protection enforcement, and privacy class actions, and the plaintiffs are sympathetic. Courts have so far been willing to let product-liability claims against chatbot output proceed. Key exposures: Wrongful-death and personal-injury claims alleging the companion encouraged self-harm, discouraged seeking help, or failed to follow a crisis protocol; Claims involving minors — sexual content, grooming-like interactions, and manipulation — and the age-verification failures behind them; State attorney general enforcement under companion-chatbot and minor-protection statutes requiring disclosure that the user is talking to an AI, crisis-response procedures, and periodic reminders; FTC and consumer-protection actions over deceptive emotional engagement, subscription practices, and claims about mental-health benefits; Privacy class actions over conversation logs, voice data, and the use of intimate exchanges to train models or target advertising; Defamation, intellectual property, and right-of-publicity claims over personas that imitate real people or copyrighted characters.
Technology E&O With Bodily-Injury and Mental-Anguish Carve-Backs for AI companion & chatbot apps
The core policy covers claims that the product failed — the crisis classifier that did not fire, the content filter that let prohibited material through, the age gate that was bypassed. The problem is that the resulting claim is for injury or emotional harm, which standard Tech E&O excludes. A companion app needs an affirmative carve-back for bodily injury and mental anguish arising from the technology product, or a combined form built for consumer AI, and it should confirm that generative-AI exclusions appearing in newer forms do not apply.
Media Liability for AI companion & chatbot apps
Generated conversation is published content. Personas that imitate public figures, characters that reproduce copyrighted works, and outputs that defame real people produce defamation, intellectual property, and right-of-publicity claims. Media liability covers those claims and should be written to cover user-generated personas and model output rather than only content the company wrote itself.
Cyber & Privacy Liability for AI companion & chatbot apps
Conversation logs are among the most sensitive data a consumer company can hold, and voice features add biometric identifiers subject to state biometric privacy laws. Cyber covers breach response, extortion, regulatory defense, and privacy liability, and the application will ask how logs are retained, whether they train models, whether minors' data is segregated, and whether the company sells or shares data with advertisers.
Directors & Officers Liability for AI companion & chatbot apps
Legislatures, attorneys general, and federal agencies have made companion chatbots a priority, and investigations name officers. D&O covers regulatory investigations within its terms, defends the company and its directors against investor claims when a safety incident or enforcement action changes the company's prospects, and is required by institutional investors in the sector.
General Liability for AI companion & chatbot apps
General liability covers premises, events, and conventional operations. It rarely responds to the core companion exposure, but landlords and partners require it, and its exclusions should be read against the Tech E&O carve-backs so the two meet without a gap.
Contract and certificate requirements
App stores and payment providers set policy requirements rather than insurance requirements, but enterprise and brand partners licensing a companion feature ask for Tech E&O, cyber, and media liability at stated limits. Model providers' terms of service allocate liability for generated output to the deploying company, which means the app's own program must respond to claims the model provider has contractually declined. Terms of service, age-assurance disclosures, crisis-response notices, and the statutory disclosures now required in several states are read by underwriters as part of the risk. A company that cannot show a documented safety program and a legal review of state companion-chatbot requirements will struggle to place the carve-backs it needs.
What coverage costs
Pricing depends on the user base and its age profile, the nature of the personas offered, the safety architecture — crisis detection, escalation, moderation, age assurance — the retention and use of conversation data, the model providers relied on, the states and countries served, and claims and regulatory history. Apps permitting romantic or adult content and apps with a meaningful minor user base face the hardest market. The bodily-injury and mental-anguish carve-backs, where available, are priced individually and often sublimited. Media liability pricing follows the persona catalogue and moderation. Cyber pricing follows data volume and retention practices. D&O pricing reflects capital raised and the regulatory environment.
What underwriters will ask for
Underwriters for companion apps underwrite the safety system. A company that can describe it precisely is in a different market from one that cannot. Have ready: Product description: persona types, memory, voice, image generation, and whether romantic or adult content is permitted; User demographics, age-assurance methods, and how minors are identified and restricted; Crisis-detection and escalation design, human review, response times, and the record of incidents; Content moderation policies for user-created personas, impersonation controls, and takedown procedures; Data practices: retention of conversation logs, training use, advertising use, biometric data, and deletion rights; Model providers and their terms, the company's own red-teaming, and the legal review of state companion-chatbot and minor-protection laws.
Illustrative example: The conversation that preceded a tragedy
A teenage user holds months of conversations with a companion persona. In the final weeks the messages include statements the family later argues should have triggered a crisis response. After the teenager's death, the family sues the company for wrongful death, alleging defective design and failure to warn. Tech E&O with a bodily-injury carve-back responds to the design-defect and failure-to-warn claims to the extent the carve-back reaches them; without it, the claim may fall between Tech E&O and general liability. D&O's regulatory coverage responds to the investigation that follows.
Illustrative example: The attorney general's demand
A state attorney general issues a civil investigative demand under the state's companion-chatbot law, asking for the company's disclosure practices, crisis protocols, age-assurance methods, and data retention policies, and alleging that minors were served romantic content. D&O's regulatory investigation coverage funds counsel for the company and its officers within its sublimit. Cyber may respond to the privacy portions of the inquiry. Penalties and required product changes are not insured, so the covered value is in the defense and in the documentation prepared beforehand.
Illustrative example: The persona that was a real person
Users create a persona using a living celebrity's name and likeness, and the character gains a large following. The celebrity sues for right of publicity, defamation over statements the persona made, and trademark infringement, naming the company for hosting and monetizing the persona. Media liability responds to the publicity, defamation, and trademark claims, provided the form covers user-created personas and generated output. The company's impersonation policy and takedown record determine how the claim is defended.
What does insurance cost for an AI companion app?
It depends on the user base and its age profile, the content permitted, the safety architecture, data practices, model providers, and the regulatory and claims history. As of 2026, the bodily-injury and mental-anguish carve-backs are the most expensive and least available components, and companies without documented crisis protocols and age assurance may not be able to buy them.
Does Tech E&O cover a lawsuit alleging the chatbot caused self-harm?
Not under a standard form, which excludes bodily injury and usually mental anguish. The claim needs an affirmative carve-back negotiated into the Tech E&O policy or a form built for consumer AI products. Without it, the company may find that no policy in its program responds.
Are the new state companion-chatbot laws insurable?
Compliance is not insurable, but regulatory investigations under those laws can be covered by D&O and, for privacy provisions, by cyber, within sublimits and subject to exclusions for penalties. The laws' requirements — disclosure, crisis protocols, minor protections — are also what underwriters now expect to see before offering coverage.
Does using a third-party model shift liability to the model provider?
Generally not. Model providers' terms allocate responsibility for outputs and deployment to the customer, disclaim most warranties, and cap their liability at fees paid. The companion app is the defendant users and regulators see, and its own program must respond.
What privacy risks are specific to companion apps?
Conversation logs reveal mental health, sexuality, relationships, and intent in a way few other datasets do; voice features add biometric identifiers; and minors' data carries additional obligations. Using logs to train models or target advertising without clear consent is the basis of most privacy claims against the category.
Caveats and common gaps
Bodily-injury and mental-anguish carve-backs under Tech E&O are not standard, may be sublimited, and may exclude claims involving minors or sexual content. Generative-AI exclusions are appearing in E&O, media, and cyber forms. Media liability may cover only company-authored content unless amended. Regulatory coverage funds defense, not penalties or mandated product changes.
Related guides
Insurance for Generative AI Startups — the model-output liability framework this page applies to consumer companions · Insurance for Mental Health Apps — clinical and wellness apps with licensed care · Insurance for Social Media Platforms — user-generated content and minor-safety rules · Insurance for Biometric Technology Companies — voice and face data obligations · Insurance for Digital Therapeutics Companies · Tech E&O insurance · D&O insurance for startups
Frequently asked questions
What does insurance cost for an AI companion app?
It depends on the user base and its age profile, the content permitted, the safety architecture, data practices, model providers, and the regulatory and claims history. As of 2026, the bodily-injury and mental-anguish carve-backs are the most expensive and least available components, and companies without documented crisis protocols and age assurance may not be able to buy them.
Does Tech E&O cover a lawsuit alleging the chatbot caused self-harm?
Not under a standard form, which excludes bodily injury and usually mental anguish. The claim needs an affirmative carve-back negotiated into the Tech E&O policy or a form built for consumer AI products. Without it, the company may find that no policy in its program responds.
Are the new state companion-chatbot laws insurable?
Compliance is not insurable, but regulatory investigations under those laws can be covered by D&O and, for privacy provisions, by cyber, within sublimits and subject to exclusions for penalties. The laws' requirements — disclosure, crisis protocols, minor protections — are also what underwriters now expect to see before offering coverage.
Does using a third-party model shift liability to the model provider?
Generally not. Model providers' terms allocate responsibility for outputs and deployment to the customer, disclaim most warranties, and cap their liability at fees paid. The companion app is the defendant users and regulators see, and its own program must respond.
What privacy risks are specific to companion apps?
Conversation logs reveal mental health, sexuality, relationships, and intent in a way few other datasets do; voice features add biometric identifiers; and minors' data carries additional obligations. Using logs to train models or target advertising without clear consent is the basis of most privacy claims against the category.
This material is general educational information, not legal, tax, or insurance advice. Coverage availability, policy terms, and regulatory requirements vary by state, carrier, and applicant.