Is Microsoft training AI models using consumers' data from Office apps?

Connected Experiences feature, which analyses user content to provide suggestions and recommendations, has raised questions about whether Microsoft is using customer data to train its AI models

Concerns regarding user data privacy have emerged following the mention of the "Connected Experiences" option in Microsoft's productivity-centric Office apps. This feature, which analyses user content to provide suggestions and recommendations, has raised questions about whether Microsoft is using customer data to train its AI models.

 

Microsoft has clarified that it does not use customer data to train its AI models. In response to a post on X (formerly Twitter) suggesting that Microsoft might be using users’ Word and Excel documents to train AI, the company confirmed that it does not use customer data from Microsoft 365 apps to train large language models (LLMs).

The concern stemmed from the privacy setting for "Connected Experiences," an optional feature within Microsoft Office. In a blog post outlining the feature, Microsoft explained that Connected Experiences analyses users' Office content to provide design recommendations, editing suggestions, data insights, and more. However, the company did not explicitly state whether user data is used for other purposes, such as training AI models. Additionally, the feature is enabled by default in Office, and users must manually toggle it off to opt out.

 

A post from the X handle of nixCraft, a forum for Linux and Unix users, raised concerns about data integrity within Microsoft Office. In response, Microsoft clarified via its official Microsoft 365 handle that, “In the M365 apps, we do not use customer data to train LLMs. This setting only enables features requiring internet access like co-authoring a document.”

 

Frank Shaw, Communications Head at Microsoft, further addressed the issue regarding Connected Experiences in a post on the social media platform Bluesky. Responding to claims that Office scrapes user documents unless they opt out, Shaw stated that such claims were untrue, adding, "Following the link for more information makes that clear."