Microsoft refuses to train AI on customer data following uproar

Microsoft Corp. has rejected claims that it uses customer data from Microsoft 365 consumer and business applications, such as Word and Excel, to train large language models, after user concerns spread on social media.

Allegations emerged a few days ago after some users noticed they were automatically enabled in Microsoft 365 Connected Experiences, a 2019 setting that allows product features to be connected to the internet and the cloud.

Allegations quickly spread among Microsoft users over what they perceived to be a “sneaky” update to the vendor’s terms of service, which they said could allow the company to scrape Microsoft Office documents to train AI models.

The parameter in question did not specifically refer to AI training.

“These claims are false,” a Microsoft spokesperson told Bloomberg Law on Wednesday, noting that some customers may consent to their data being used in “certain cases,” such as for the development of custom models when “this is appropriate.” explicitly requested.”

The Connected Experiences setting referenced has “no connection to how Microsoft trains large, foundational language models,” the spokesperson said. This setting enables cloud-based features like real-time co-authoring, cloud storage, and tools like the editor in Word that provide spelling and grammar suggestions, they said.

Users of online platforms are increasingly concerned that their data, whether social media posts, contracts or photos, is used to train organizations’ latest AI models .

This summer, Adobe Systems Inc. confronted a similar tumult when it updated its terms of service, which users claimed allowed Adobe to seize their intellectual property and use their data to power AI models. The Photoshop and Illustrator maker released updated terms of service a few days later, in which it specifically stated that the company does not train its Firefly AI models on its cloud content.