Google has denied the rumors claiming that Gmail messages were used to train its Gemini AI, clarifying that the AI is trained on a completely separate dataset. The confusion began with a misinterpretation related to Gmail’s Smart Features, which some thought implied that user emails were used for AI training. Google spokesperson Jenny Thomson stated that Gmail messages are not included in training Gemini and that Smart Features are opt-in, with no settings changed without user consent.
Google emphasized that Gemini’s training data does not include personal Gmail content. MalwareBytes initially published a story fueling the confusion but later updated it to clarify the misunderstanding, noting that Gmail scans messages only to enable built-in features and not for training generative AI models. Google assures users that those who do not enable Smart Features have their data excluded from any such processing.
This clear response from Google aims to reassure users about their privacy concerning Gmail data and AI training. Gemini AI training relies on separate, publicly accessible data sources and not on private Gmail messages, maintaining user trust and data protection.