WIRED Says Your Chatbot Secrets Are Not Yours. Zero Retention Contracts Are The Only Real Fix.
WIRED's privacy guide establishes a baseline expectation of approximately zero real privacy when using ChatGPT, Claude, or Gemini, assuming anyone with legal access can obtain your conversation records. The only robust exception identified is a zero data retention or ZDR contract between your employer and the AI provider that legally prevents record retention. Proton, a privacy-focused cloud company, offers an AI chatbot built around this contractual protection.
This story illustrates the principle of contractual privacy enforcement, which I must reluctantly explain since the alternatives are apparently too difficult for people to grasp. The mechanism at work is that privacy guarantees from AI providers are not technological but legal. You cannot encrypt your way out of a subpoena. You need a contract that forbids data retention in the first place, which is a distinction most users never consider.
WIRED published the guide. Proton offers a privacy-focused AI chatbot implementing the ZDR approach described.
- Open ChatGPT, Claude, or Gemini and type a test message containing personal information you would not want subpoenaed. Observe that the provider retains this by default, because of course they do.
- Visit Proton's website and review their AI chatbot offering, which is built around zero data retention contracts. Compare the privacy terms to the big three providers.
- If you have an employer, ask your IT department whether your organization has a ZDR contract with any AI provider. If yes, use that tool for sensitive queries. If no, assume everything you type can be read by someone with a legal path to it.