Gmail said in an announcement that Google does not train its foundational artificial intelligence models, including Gemini, on users' personal emails.
The announcement answers privacy questions as new AI features are added inside Gmail and clarifies how Gemini interacts with users' inboxes.
Related reading
- Anthropic cuts off Claude Code subscribers from third-party AI agents, starting with OpenClaw
- Caltech spin-out PrismML launches 1-bit AI model that fits into 1.15 GB and runs on Apple devices
- Anthropic spots 'emotion vectors' inside Claude
Google said any access granted to Gemini in Gmail is limited to isolated tasks, such as summarizing a lengthy email, and is used only to complete that request. The company added that Gemini only processes the information needed for the specified task and then leaves the inbox.
The announcement also states, "Your inbox is your business." Google directed readers to learn more from "Gmail’s VP of product, Blake Barnes" for additional details about how Gemini and Gmail work together.
The recap
- Gmail says Gemini will not be trained on emails.
- Gemini processes only requested tasks inside the user's inbox.
- Google points readers to Gmail’s VP of product for details.
