How to Use AI With Your Privacy Intact
WIRED guide examines AI chatbot privacy risks and zero-data-retention caveats, highlighting Moxie Marlinspike's cryptographically private Confer chatbot.
The article details how ChatGPT, Claude, and Gemini default to storing highly personal conversations that can be shared, used for training, or obtained via legal process. Enterprise zero-data-retention contracts from OpenAI, Anthropic, and Google delete records after processing but carry exceptions: Anthropic excludes Mythos-class models like Fable 5.1, OpenAI's Private Safety Processing analyzes activity before deletion, and Google logs sanitized prompts. Moxie Marlinspike's new Confer chatbot uses cryptography to prevent its own servers from logging conversations, unlike policy-based promises such as Proton's Lumo. Consumer options generally offer weaker privacy guarantees than enterprise ZDR.
- Major chatbots default to storing conversations accessible to legal process and partners
- Enterprise ZDR contracts delete records but include abuse-monitoring exceptions
- Anthropic excludes ZDR for Mythos-class models like Fable 5.1
- Marlinspike's Confer applies cryptography to block server-side conversation logging
- Proton Lumo and similar consumer tools rely on no-log policies, not technical guarantees
From the original article
10 min read at wired.com
Your conversations with AI chatbots are both highly personal and deeply vulnerable to surveillance. Here’s how you can protect yourself.
By Andy Greenberg · wired.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.