To train their artificial intelligence models, AI laboratories need immense amounts of data. For example, if you want to train an AI that generates text, you need to have a significant amount of data that includes text. And recently, rumors accusing Microsoft of using Microsoft 365 user data to train AI models have circulated online.
These accusations are serious because people who use Microsoft 365 products, like Word or Excel, may use them to write confidential documents that should not be exploited by an AI. The Redmond firm has already denied these rumorsin a publication on X.
“In M365 applications, we do not use customer data to train LLMs”assures the company. This also explains that the “connected experience” menu which was singled out by the rumor “only enables functions requiring internet access, such as co-writing a document.”
A big misunderstanding
As the How To Geek site notes, it is possible that the panic among some users is partly due to the language used by Microsoft, which is not clear enough. In the settings of its Office applications, such as Word, it has set up a menu which allows you to activate or deactivate the famous “connected experiences”. And one of the options, enabled by default, mentions “connected experiences that analyze your content”.
At a time when data is essential to evolving artificial intelligence models, this may be enough to panic some users. In any case, Microsoft also confirmed to the How To Geek site that it “does not use customer data from Microsoft 365 business and consumer apps to train large language models.” Furthermore, the parameters of connected experiences “have nothing to do with how Microsoft trains large language models.”
- Microsoft denies rumors accusing it of using user content from apps like Word or Excel to train artificial intelligence models
- The settings that are singled out by these rumors only allow access to features that require an internet connection
- They have no link with the way the Redmond firm develops its artificial intelligence models
