Google denies secretly training AI on Gmail users' emails
Google has been accused of implementing features that give its artificial intelligence models access to users' private messages and files, a claim Google officials have called «misleading».
Google has been accused of implementing features that give its artificial intelligence models access to users' private messages and files, a claim Google officials have called «misleading».
Last week, cybersecurity firm Malwarebytes warned in a blog post that Google was making changes to Gmail users that would allow the company to view their private emails and files to train Gemini and other AI tools. Among the features that gave this access were Smart Compose, Smart Reply, and text prediction.
Google called this statement «misleading».
«These messages are misleading — we have not changed any user’s settings. Gmail’s intelligent features have been around for many years, and we do not use the content of your Gmail to train our Gemini AI model. Finally, we always communicate openly and clearly about changes to our terms of service and policies,» the company representative noted.
Malwarebytes later added a significant correction to the story, stating that «the way Google recently rewrote and released [these features] led many people (including us) to believe that Gmail content could be used to train Google’s AI models, and that users were automatically opting into the feature.»
The company admitted that «after a thorough study of the documents» this information is not true.
Last week, Google began rolling out a new artificial intelligence model, Gemini 3, which it called the «most accurate» and «smartest» in the world and a big step towards AGI.
Artificial intelligence expert and author of a number of books and courses on AI, Oleksandr Krakovetsky, evaluating Google’s new Gemini 3 model, called it a breakthrough based on benchmarks. At the same time, Krakovetsky noted that he still feels a significant difference between benchmarks and real-world use.
Read the country’s main IT news in our Telegram
Read the country’s main IT news in our Telegram
A critical bug was found in ChatGPT: attackers could steal data from Gmail using a single email
A critical bug was found in ChatGPT: attackers could steal data from Gmail using a single email
Google unveiled its «smartest AI model» Gemini 3, which outperforms GPT-5 Pro and Claude Sonnet 4.5
Google unveiled its «smartest AI model» Gemini 3, which outperforms GPT-5 Pro and Claude Sonnet 4.5
«The Gemini paradox is that in real life it doesn’t perform very well compared to Claude and ChatGPT. And I don’t understand why, given these figures.» Experts on Google’s new Gemini 3 model
«The Gemini paradox is that in real life it doesn’t perform very well compared to Claude and ChatGPT. And I don’t understand why, given these figures.» Experts on Google’s new Gemini 3 model
