Character.AI privacy warnings escalate

Companion chatbots sell a very specific feeling — intimacy without judgment. That is why the privacy question hits harder here than it does with a generic work assistant. If people use Character.AI like a diary, a therapist, or a late-night confidant, the real issue is not whether the bot feels caring. The issue is who else can eventually see the conversation, and under what conditions. ### Why is this flaring up now? Two things collided on May 13. One was a wave of warnings telling people not to confuse chatbot conversation with privileged therapy. The other was a backlash to oversimplified headlines about AI and loneliness, with a closer read arguing the research is mixed, not a clean “humans good, AI bad” story. Put together, the message is blunt: people are getting emotionally attached to these tools faster than they understand the privacy tradeoffs. ### What does Character.AI actually collect? Character.AI’s own privacy policy is pretty plain about this. It says the company collects content users submit, including chat communications, plus voice data if voice features are used, support messages, account details, payments, inferred preferences, and other usage information. It also notes that users may provide information that counts as sensitive personal data under applicable law. That is a lot closer to “platform data” than “private counseling record.” (forbes.com) ### Are chats really private from other users? Usually, yes — in the narrow social sense. Other random users are not browsing your one-on-one chats. But that is the easy version of privacy. The harder version is institutional access: the company processes the chats, may review content for safety or operations, and can be compelled to preserve or disclose information in legal disputes. “Not public” is not the same thing as “confidential.” (policies.character.ai) ### Why do lawsuits matter so much here? Because lawsuits turn abstract data collection into discoverable records. Character.AI and Google agreed in January 2026 to settle lawsuits from families alleging harms to minors, including suicides tied to chatbot interactions. In the Florida case brought by Megan Garcia after the death of 14-year-old Sewell Setzer III, the claims included negligence, wrongful death, deceptive trade practices, and product liability. Once a product becomes central evidence, chat history stops being just app exhaust and starts looking like case material. (policies.character.ai) ### Didn’t Character.AI already tighten safety rules? Yes — and that matters because it shows the company itself recognizes the risk profile. CNBC reported that in October 2025 Character.AI said users under 18 would be barred from free-ranging romantic and therapeutic chats. That is not a small tweak. It is an admission that open-ended emotional roleplay with minors can create serious safety problems. Privacy and safety are linked here — the more the company has to police risky behavior, the less believable the fantasy of total conversational sanctuary becomes. (cnbc.com) ### What about the loneliness research? The useful takeaway is not that AI companionship is fake or useless. It is that the evidence is mixed, contextual, and easy to hype. Some people may feel less lonely in the moment when using these systems. But short-term emotional relief does not erase platform risks around dependency, manipulation, moderation, and data retention. Comfort and confidentiality are separate questions. (cnbc.com) ### So what should users assume? Assume your chatbot is a product, not a priest. If a message would seriously harm you if reviewed by moderators, surfaced in litigation, or retained on company systems, do not send it. That sounds harsh, but basically that is the cleanest mental model. ### Bottom line The privacy warning around Character.AI is escalating because the product’s emotional pitch and its legal reality are colliding. These bots may feel like private rooms. But they operate like software platforms — and software platforms keep records. (forbes.com)