Google Denies Rumors: Gmail Messages NOT Used to Train Gemini AI Models

Google has denied social-media rumours that Gmail messages and attachments are being used to train Gemini AI models, insisting that no policy changes have occurred and that such reports are misleading. The firm emphasises that Gmail Smart Features rely on automated processing for conveniences like spell-check and calendar integrations, and that enabling them does not authorise transferring email content for AI training.

A Google spokesperson, Jenny Thomson, is quoted in the piece as saying these reports are misleading and that Google does not use Gmail content to train Gemini. The article notes that the consent language within Google Workspace relates to improving collaboration and usability rather than authorising AI training, and explains a January update separated personalization settings between Google Workspace and other products.

While some reports suggested previously disabled smart features reappeared, the piece frames this as not indicative of a policy shift, and advises privacy‑conscious users to review their account settings. According to The Verge, these statements underscore Google’s position that no data fed into AI training is occurring under current settings.