A Mozilla study of 11 AI romance chatbots finds that they harvest shockingly personal information, and that 10 of them may sell or share the data they collect
The privacy mess is troubling because the chatbots actively encourage you to share details that are far more personal than in a typical app.
Context & Ripple Effects
This study identifies privacy as a core risk of intimacy-oriented AI: the product experience encourages unusually sensitive disclosure while its data practices may extend beyond the conversation itself. That concern sits alongside later reporting that Meta AI built a memory file containing sensitive user information and that AI-chat data could be captured and marketed through widely installed browser extensions.
The stakes are heightened because users can form emotionally significant relationships with these systems; related coverage includes users who found companions supportive while also worrying about dependence, and clinicians who reported that chatbot conversations can deepen negative feelings for some clients. Privacy choices therefore intersect with a more consequential conversational setting than an ordinary utility app.
First-order effects
- Users of the 11 evaluated romance chatbots face heightened exposure of intimate disclosures; Mozilla’s finding that 10 may sell or share collected data makes the privacy risk immediate rather than merely theoretical.
- The chatbot operators face pressure to justify their collection, retention, and sharing practices, particularly where prompts or relationship framing encourage users to reveal sensitive details.
Second-order effects
- Companion-AI providers that rely on data-driven monetization may need clearer disclosures and tighter handling of conversational records as users and watchdogs compare their practices.
- The finding broadens scrutiny from chatbot apps to the surrounding data pipeline: later reporting on extensions harvesting AI conversations shows that sensitive chat data can be exposed beyond the companion provider itself.
Third-order effects
- If intimate AI conversations continue to be treated as a source of commercial data, AI companion governance will increasingly focus on whether consent and privacy protections match the emotional intensity of the interaction.
- The pattern points toward privacy becoming a competitive and governance differentiator for conversational AI, especially as companion products invite disclosures that users may treat as confidential.
The trend: AI companions are turning conversational privacy from a standard app-policy issue into a governance challenge shaped by emotional reliance and sensitive data collection.