HOT 285: Are My AI Chatbot Conversations Private? - The Risks Of Sharing Information with AI Chatbots
Summary
Micah Sargent answers a listener’s question about AI chatbot privacy, confirming that conversations are stored on company servers and used for model training by default. He details how to disable training in ChatGPT and Gemini, noting that Gemini disables chat history when training is off. The episode explains that human reviewers can access chats, there is no HIPAA or legal privilege, and deleted data may persist. Sargent advises using dedicated health tools or temporary chats for sensitive information and removing personal identifiers before uploading documents.
Topics discussed
Sponsor: Trusted Tech Microsoft 365 licensing and support
Introduction and Wendell's question about AI chatbot privacy
Short answer: Chatbots store data and train models by default
Where data goes: Server storage, account ties, and training defaults
Work vs. personal accounts and the trade-off with Gemini history
Who can see your chats: Human reviewers and lack of legal privilege
Legal holds, the NYT lawsuit, and advertising data usage
Chatbots vs. Google Search: Data volume and persistent memories
HIPAA does not apply to AI chatbots or user-typed data
ChatGPT Health: A dedicated, non-training space for medical data
Advice: Redact personal identifiers before uploading lab results
How to turn off training and manage settings in ChatGPT
Gemini settings: Activity, auto-delete, and temporary chat limitations
Verdict and quick-fix summary for protecting your privacy
Listen ad-free on Castria