Lawsuit alleges LinkedIn shared DMs for third-party AI training

How training data for AI is sourced has been a point of contention for as long as companies have been training AI models. Corporations assert that training AI on publicly accessible information on the internet constitutes fair use of that information, because the models that incorporate data parsed from YouTube videos or blogs, for example, are transformative — in other words, that they fundamentally change the data they’ve absorbed before redistributing it, creating legally distinct works. Whether that argument can withstand long-term legal scrutiny remains to be seen.

Microsoft-owned LinkedIn has found itself at the center of a different kind of AI training data controversy. As reported by the BBC (and spotted by TechRadar), a California lawsuit accuses LinkedIn of sharing private user data — including user-to-user direct messages — with third parties as fodder for AI training, without adequately notifying users or giving them the opportunity to opt out of the arrangement.

According to the suit, LinkedIn “quietly” introduced a new privacy setting that automatically opted users into a program that shares user data with third parties for the purposes of training AI. The suit also alleges that LinkedIn updated its FAQ section to say that users had the option not to share data in this way, but also that opting out wouldn’t have any effect on data that had already been shared.

LinkedIn denies any wrongdoing

The suit alleges that LinkedIn’s policies described here were in violation of the Stored Communications Act. It’s seeking damages of $1,000 for each user affected.

In part, the lawsuit reads that LinkedIn’s actions “indicate a pattern of attempting to cover its tracks.” For LinkedIn’s part, a company spokesperson told BBC that the suit’s assertions “are false claims with no merit.”