Speculation has been rampant on social media, suggesting that Google is covertly using content from Gmail to train its Gemini AI model. Recently, viral posts claimed that emails, attachments, and personal conversations were being accessed for AI training, further fueled by allegations that Google modified privacy settings without informing users. One user highlighted a supposed automatic opt-in feature that allows Gmail to utilize private messages and attachments for AI training, urging others to disable Smart Features in two different settings. In response to these rumors, Google spokesperson Jenny Thomson provided a statement to a prominent publication, emphasizing that these claims are unfounded.
She stated that the fears circulating online are misinformed, asserting, “These reports are misleading – we have not changed anyone’s settings, Gmail Smart Features have existed for many years, and we do not use your Gmail content for training our Gemini AI model.” Despite this clarification, the uproar has led many Gmail users to reassess their settings, uncertain about any actual changes. The confusion regarding Smart Features appears to stem from an update Google implemented earlier this year, which allowed users to manage these features independently across its services. This means that options in Gmail or Workspace, such as auto-draft suggestions and email categorization, can be turned off without impacting similar features in other applications.
However, reports suggest that some users found their previously disabled settings were reactivated without their knowledge. While the extent of this issue remains unclear, it has raised concerns that Google might be re-enabling features to enhance its AI data sources. These tools go beyond simple spell checks and can automate tasks like flight updates and summarizing emails. When users activate these features, Gmail notifies them that the service will utilize “Workspace content and activity to personalize your experience.” Although this may raise alarm for some, Google asserts that such personalization remains confined to the user’s account and does not contribute data to the training of Gemini.
The increased scrutiny coincides with Google’s promotion of Gemini 3, its latest and most sophisticated AI model. This system is designed to understand and analyze information with what the company refers to as human-like depth. CEO Sundar Pichai emphasized Gemini 3’s improved long-context reasoning and its capability to analyze diverse content ranging from academic papers to video lectures. He stated that this new model is “state-of-the-art in reasoning, built to grasp depth and nuance.” Pichai also noted that Gemini 3 is more adept at understanding the context and intent behind user requests, minimizing the need for extensive prompting. As Google advances the capabilities of AI, concerns regarding the use of personal data are likely to increase.
For now, Google maintains that user inboxes do not contribute to its AI advancements, yet the ongoing debate illustrates the vigilance with which users are monitoring updates.