Do Word and Excel collect data for AI training?

Microsoft recently introduced a new standard feature in Office 365 that is generating excitement across the industry. “Connected experiences” allow the company to automatically download and process data from applications such as Word and Excel. Officially, this aims to help users create, communicate and collaborate more effectively. But experts are concerned about data protection and the possible use of the collected information for AI training. The new feature is enabled by default, but opting out is deeply hidden in Office 365’s privacy settings, as a security researcher named @nixCraft on Twitter/X. Unless users explicitly disable the option, their document data will be collected and processed. This has led to speculation about whether Microsoft could use this data to train AI models like Copilot.

Microsoft itself remains vague about what exactly is happening with the data it collects. A statement simply says: “Connected experiences that analyze your content are experiences that use your Office content to provide you with design suggestions, edit suggestions, data insights, and similar functionality. » There is no explicit mention of AI training. However, the Microsoft Services Agreement contains a clause that grants the company a “worldwide, royalty-free intellectual property license to use your content.” This could theoretically also include training AI models, although Microsoft has so far denied this. In any case, a Microsoft spokesperson explicitly denied that M365 data is used to train AI models, telling The Stack: “In Microsoft 365 consumer and business applications, Microsoft does not use customer data to train large language models without their permission.” The question is whether a default setting – from Microsoft’s perspective – is ever considered a permission.