AI chatbots are increasingly used for highly personal conversations, yet their default settings often allow data to be stored, shared, used for training, or disclosed through legal demands. Privacy experts warn that these systems can create exceptionally detailed profiles of users, making AI interactions a major emerging surveillance risk.

The article highlights privacy-focused alternatives such as Signal creator Moxie Marlinspike’s Confer, which uses cryptography to prevent its servers from logging conversations. It also discusses “zero data retention” agreements offered to enterprise customers by OpenAI, Anthropic and Google, while noting that these policies are generally unavailable to ordinary users and may contain important exceptions.