Meta recently revealed that it was using its employees as a source of training data, essentially tracking everything the humans were doing so an AI could learn from (or replicate) it. A new report from Business Insider claims that the social media company has since pulled back from this initiative.
Why the change? It turns out that if you’re making a record of everything your company does, then somewhere on your systems is… a record of everything your company does. Nobody wants that sort of information leaking out, whether it’s to competitors or to the human resources department.
Meta, mind?
In this case, Meta put the brakes on its program after sensitive information became available across the company. Information, including private conversations, transcripts, and performance data, was visible to folks who really shouldn’t have had access. Whether they (whoever ‘they’ are) accessed it isn’t known at this point.
According to the report, a company spokesperson said, “We have carefully designed this program with privacy safeguards, and while we have no indication at this time that Meta employees improperly accessed any data, we’re pausing it while we investigate.”
The program is (or was) known as the Model Capability Initiative (MCI). Other outlets, including Reuters, claim it’s still running internally at Meta. The MCI’s reach is broad enough that a shutdown will have to be rolled out in phases.
