Microsoft has addressed the controversy that arose a few weeks ago regarding the alleged use of users’ Microsoft 365 documents (like Word and Excel) to train AI models, such as Copilot. The company stated that it is not using any customer data from its Office services for this purpose and clarified that the confusion stems from a security feature.
Please follow us on Facebook and Twitter.
Specifically, users have been voicing concerns for weeks about an option in Microsoft 365 that appeared to ask them to opt out of training AI models. This gave the impression that data collection for developing language models was enabled by default.
Microsoft has explained that the option users are concerned about is actually related to features that let them “search for images online” or “find information available online.” These features are part of what the company calls “optional connected experiences.” This option is enabled by default but can be turned off. Additionally, Microsoft has stated that there is no reference to using customer data for AI training in this context.
However, many users argue that the wording of this option suggests that the features “analyze your content” and does not explicitly rule out the possibility of using data for AI training.
No, Microsoft Doesn’t Use Your Word Data to Train Its AI.
In response to the complaints, Microsoft reiterated in a post on X (formerly Twitter) that Microsoft 365 apps do not collect data to train artificial intelligence models. “This setting only enables functions that require internet access, such as co-authoring a document,” the company explained. Frank Shaw, Microsoft’s chief communications officer, also clarified on Bluesky that the claims about using data from platforms like Word or Excel for AI training are false.
Microsoft’s situation is similar to Adobe’s case, where the platform’s terms initially implied that user data might be used to train AI models. Adobe addressed the controversy by updating its terms with clearer language, explicitly stating that it does not use images or other user content for AI training purposes.
In any case, with the growing use of AI, many platforms do collect data to train language models. However, all of them, especially within the European Union, are required to provide users with an option to decide whether they want their data to be collected or not.