Microsoft denies rumors and uses its documents to train AI

Microsoft explains: Microsoft 365 suite user documentation is not used to train large language models (LLMs). The company took a position on the issue After a tweet about data collection went viral on X, formerly Twitter.

“Beware: Microsoft Office and several other companies have in recent months maliciously enabled an optional feature that scans your Word and Excel documents to train internal AI models.” Original post on X by user @nixcraft said. “This option is enabled by default and you must manually uncheck it to disable it,” he continued.

In the post, the user highlighted the “Connected Experiences” function in the document settings menu in Microsoft 365 applications. Data collection will take place through this tool.

In M365 applications, we do not use customer data to train graduates. This setting only enables features that require internet access, such as co-authoring a document. https://t.co/o9DGn9QnHb

— Microsoft 365 (@Microsoft365) 25 November 2024

The original post had 536.9k views at the time of writing. Probably, This was enough to catch Microsoft’s attention In one of the answers he reinforced that personal documents are not used in Master’s studies.

“In Microsoft 365 applications, we do not use user data to train LLMs. This configuration only enables features that require an internet connection, such as co-authoring of documents,” he noted.

Wrong but needs attention

Even though the warning the user posted is not true, strengthens users’ vigilance regarding data collectionmainly for tools aimed at AI training. Considering Microsoft’s controversial history in the region, it’s no surprise that a rumor like this is floating around the internet.

In any case, this is not the case. So, if you are a professional who uses applications from the Microsoft suite (e.g. Word, Excel or PowerPoint) to produce completely custom content, no need to worry about scanning your documents to train company models.