Google clarified how Gmail data interacts with its Gemini AI models.

Gemini training with Gmail rumors debunked.
Recently, posts on X (formerly Twitter) and Threads have warned that Google is using Gmail data to train AI models. These posts reference the "Smart features and personalization" setting in Gmail and advise users to disable it. The claim gained traction, and some tech security news sources initially reported it before issuing corrections.
Google has addressed these concerns, stating that Gemini is not trained using Gmail content. They clarified that the "Smart features" setting, which enables functions like tab sorting and Smart Compose, has existed for some time and is not a new feature designed to collect data for LLMs. The issue appears to be a case of online misinformation.
This situation highlights concerns about AI and data privacy. Companies like OpenAI have faced scrutiny over data usage, causing user vigilance. Apple promotes its upcoming "Apple Intelligence" as a privacy-focused alternative, processing data on-device to avoid cloud-snooping concerns.
For Google, it's crucial to balance its AI efforts with user trust. They need large amounts of data to compete with GPT-4, but cannot afford to lose Workspace users. If users believe their emails are being used to train chatbots, they may seek alternatives. This incident shows the importance of clear communication about "smart" settings.
Google's quick response was helpful. People's concerns are understandable, considering tech companies' track records. Many use Gmail’s smart features daily, relying on the priority inbox. The confirmation that these features don't feed emails into Gemini is reassuring.
However, this should prompt Google's UI team to review its settings descriptions. If a long-standing setting can be portrayed as "spyware" in a viral post, an updated description may be needed.