Data collection for Microsoft AI training is so messy that even Redmond exec doesn't trust it

Microsoft has recently updated LinkedIn’s Term of Service (ToS), dated September 18, 2024, which, according to the document, automatically opted you into using your data for training generative AI models.

And, given Microsoft’s not-so-good track record in data collection (remember Copilot+ Recall’s PR disaster?), this update sure does raise a lot of eyebrows, even from Microsoft’s inside as well.

Scott Hanselman, a Microsoft VP for Developer Community, openly criticized LinkedIn for automatically opting users into data collection for AI training, saying it should be off by default.

“Absolutely not. And ON by default? I’ve turned this off,” he criticizes, saying that you can turn this off by going to Settings > Privacy.” Then, you can toggle off the “Use my data for training content creation AI models.”

“This setting controls the training of generative AI models used to create content. When this setting is on LinkedIn and its affiliates may use your personal data and content you create on LinkedIn for that purpose,” the toggle’s description reads.

While you can opt out by changing settings in your account, this only prevents future use of their data; any training already completed remains unaffected. It also removes personal data from training sets and does not train models on users in the EU, EEA, or Switzerland, as a part of Europe’s General Data Protection Regulation (GDPR) law.

There has been a lot of discussion on Microsoft’s AI training procedure, which surely does take a lot of money, energy, and resources. The Redmond tech giant—alongside Google— consumed 24 TWh of electricity last year for data centers alone, which are essentials in AI operations.

In other words, AI in Microsoft is massive. So much so that the company has slapped the tech into every single product that it owns, whether it’s the Windows operating system with Copilot or even social media for professionals like LinkedIn.