How Microsoft uses your Word and Excel documents to train its AI models

The Redmond firm has discreetly added a feature called “Connected experiences” which analyzes your content. “Connected experiences that analyze your content are experiences that use your Office content to provide you with design recommendations, editing suggestions, data insights, and similar features. For example, PowerPoint Designer or Translator,” indicates Microsoft on the dedicated page.

Clearly, understand that the giant sucks up content from its users to feed its large language models, and, dare we say it, possibly those of OpenAI. The two companies having a long-standing agreement, regularly fueled by large wads of dollars, it is a safe bet that Microsoft acts as a data conduit to the start-up so that it can train its models on these datasets. Would this really be a surprise?

A very difficult parameter to find, a box to uncheck

The question that remains is how to get rid of it. This is what nixCraft, author of the site Cyberciti.biz, tries to explain in a series of posts on X. “This setting was enabled by default without warning and you must manually uncheck a box to unsubscribe. If you use MS Word to write proprietary content (blog posts, novels, or other work that you intend to copyright copyright and/or sell), you will want to disable this feature immediately”, he assures.

On Windows, the steps to follow are as follows from an Office application: File > Options > Trust Center > Trust Center Settings > Privacy Options > Privacy Settings > Optional Connected Experiences and unchecking the box. All in all, seven steps to disable a critical feature that is automatically enabled.

A clause that gives all rights to Microsoft

As noted by our colleagues from Tom’s HardwareMicrosoft appears to be relying on a clause in its services contract that grants the company “a worldwide, royalty-free intellectual property license to use your content”.

In particular, we can read this: “To the extent necessary to provide the Services to you and others, to protect you and the Services, and to improve Microsoft products and services, you grant Microsoft a worldwide, royalty-free intellectual property license to use your Content, for example, to make copies, store, transmit, reformat, display and distribute via communication tools your Content on the Services.

And we can be sure that many companies have not read this clause. As a result, Microsoft potentially leverages content from its users who subscribe to Office 365, which represents approximately 345 million people worldwide.

Who is affected?

It appears that only the most recent versions (Microsoft 365) have this setting enabled by default. On its dedicated page, Microsoft explains that when users are signed in with a work or school account, these connected experiences are optional connected experiences and are provided under the terms of the Microsoft Services Agreement and privacy statement, and others Conditions may also apply.

These connected experiences in Word, Excel, PowerPoint, and Outlook (classic) from Microsoft 365 Apps will not be available to users in the following scenarios on Windows devices: Using the PowerShell BlockContentAnalysisServices advanced setting with Microsoft Purview Information Protection sensitivity labels and using dual-key encryption (DKE) to protect files and emails. One point remains to be clarified: from which version is this parameter added? For the moment, this seems difficult to determine.