Microsoft Denies Claims That Office 365 Apps Train AI Models on User Data

Microsoft on Wednesday denied claims that it uses customer data from its Microsoft 365 applications, including Word and Excel, to train artificial intelligence models. The response came after some users pointed out on social media that the company requires users to opt out of its "connected experiences" -- a feature that allows users to search for pictures or find information online, which they argued was used to train AI models.

"These claims are untrue. Microsoft does not use customer data from Microsoft 365 consumer and commercial applications to train foundational large language models," a Microsoft spokesperson said in an emailed statement to Reuters. The spokesperson added that the "connected experiences" enables features such as co-authoring and cloud storage, and has no connection to how the company trains its large language models.

Several users highlighted that the toggle for this feature is turned on by default, but there is no mention of the use of data for training AI models in the disclosure. Moreover, Microsoft's document, which the company posted in October, also fails to explain whether the user data is fed to AI models. It simply says that the "connected experiences" feature can "analyse your content" without mentioning anything about training the company's Large Language Models (LLMs).

The accusations against Microsoft -- and a subsequent rebuttal to them -- come months after Adobe was criticised for updating its terms of service, which were inferred as favouring AI training on user data. The company then clarified that it was not training its generative AI models on user data.

Companies such as Meta, X, and Google -- which are more often subjected to backlashes around user privacy and remain under the scanner -- also opt users into AI training on their data by default. It was after the users spotted these default settings in their apps that they knew their data was being used to train AI models. The conversations on social media indicate that people remain concerned about their data being used to train AI models without permission.

– Written with inputs from Reuters