Microsoft does not use Office documents to train its AI

Microsoft does not use Office documents to train its AI
Microsoft does not use Office documents to train its AI

(False) information has spread like wildfire on social networks: Microsoft would exploit users’ Office documents to train its artificial intelligence models. This is not at all the case, asserts the editor.

The rumor had the scent of credibility. On X (formerly Twitter), nixCraft, a widely followed account specializing in Unix systems, claimed thatOffice had ” discreetly » enabled a function that analyzes Word and Excel documents to train AI models for free Microsoft. An option “ enabled by default » to uncheck in an obscure corner of the office suite settings.

The fear of looting

The assertion is enough to alarm users of Microsoft software, in particular those who make their living from their pen: the publisher would therefore unduly plunder their writings for the greater benefit of its artificial intelligence, all without paying a kopeck. But it turns out that nixCraft mistook bladders for lanterns. Microsoft quickly denied this. explaining that its software did not use user data to train large language models (LLM).

The setting discussed here only concerns access to functions that require internet access, such as jointly editing a document or searching for images online. The option is indeed activated by default, but it makes no reference to AI and for good reason, it is not discussed here.

In any case, this affair shows the growing distrust of users of creation software. Unfortunately, there are many precedents where AI companies shamelessly draw on content produced by others, without authorization, for their own benefit. Adobe got its feet wet a few weeks ago on the same subject.

Read The creative community angry at Adobe’s new legal notices

The case of Microsoft is even more convoluted: the company does not just publish office software, it is also involved in generative AI with its colossal investments in OpenAI, which the training of LLMs needs maximum data. It is therefore easy to think that Microsoft could be tempted to collect documents from Office users… The burden of proof rests on these companies, and we cannot say that they have been very transparent until now.

???? To not miss any news from 01net, follow us on Google News and WhatsApp.

Source :

The Verge

-

-

PREV a major step forward in connecting AI to business data
NEXT This is what the dinosaurs ate