Training artificial intelligence is a demanding and difficult process for many reasons. On the one hand, it requires a lot of time and electricity, but at the same time huge amounts of data that a given model will be able to use for appropriate processing and learning.
This last element turns out to be extremely problematic for many corporations, forcing them to resort to illegal practices. For this reason, from time to time we hear about the use of data to which a given company did not have the right. This, of course, arouses the righteous indignation of many ordinary people who want regulators to put an end to this practice.
It turns out that Microsoft now has to face such accusations. The allegations are very serious and although the company denies them, it causes considerable distaste. What exactly is this about?
Office data used for training?
Microsoft could be in serious trouble! The company reportedly trains AI with data from the Office suite
According to the author of the Cyberciti.biz website, nicknamed nixCraft, one of the functions of the MS Office suite may illegally collect user data and then use it to train artificial intelligence. I’m talking specifically about the Connected Experiences option, which allows several users to work on documents in real time via the Internet. It turns out that the option is enabled by default.
This may mean that all documents, spreadsheets and data entered into them can potentially be processed by the service. This, in turn, leads to situations in which Microsoft may collect them and provide them for artificial intelligence training, e.g. for the Copilot program.
From the perspective of European users, it is intriguing that this feature is enabled by default. This conflicts with the provisions of the GDPR, which require manufacturers to disable options that process user data by default.
Additionally, the author of the text claims that the manufacturer of the Windows operating system does not adequately inform its users about the rights they are entitled to, much less about the actions the company undertakes.
Microsoft denies, but where is the truth?
Why does Microsoft need this data? By feeding its language models, the company can ensure the best possible quality of learning content, which is also almost unlimited. This gives it a huge advantage over competitors’ solutions.
Of course, Microsoft firmly denies this information and emphasizes that it does not train its models on the basis of data from Office or Microsoft 365 programs. The mentioned function is enabled, but only because of the possibility of co-creating documents described above.
What is the truth in the end? Of course, it is impossible to say clearly and there is little chance that Microsoft will suddenly admit to bending the law. On the other hand, there is no hard evidence that would confirm guilt and it seems impossible to prove it.
However, this situation shows how much of a problem the issue of privacy has recently become, especially when it comes to AI training. Even if Microsoft is blameless here, other corporations will be very happy to use our data and our achievements to train AI.
Can this be prevented? Of course, it is enough not to use programs and applications of specific companies. However, then we deprive ourselves of certain functionalities. So this is a very stalemate situation…
