Google is rolling out updated privacy controls for its Gemini AI app that decouple chat history retention from model training participation. Previously, users could only prevent Google from training on their Gemini interactions — including prompts, responses, uploaded files, tool calls, location data, and connected app data — by disabling chat history entirely, which erased the ability to resume past conversations. The new settings introduce independent toggles: one for chat history, which retains an adjustable retention limit, and another for AI training, which defaults to opt-out and must be manually disabled via the Activity section in Gemini settings or through the Google Account dashboard under Data and Privacy > My Activity > Other activity > Gemini Apps Activity. This change applies prospectively; Google states that data already submitted may be retained for up to three years for model improvement, and while the company attempts to filter personal information from training datasets, it offers no guarantee. The update addresses a longstanding friction between data privacy and usability in Google's AI ecosystem, allowing users to maintain conversation continuity without implicitly consenting to model training.

Read original