This won’t make the already-controversial AI training endeavor any more popular.
Meta has paused use of an AI training program that tracks its own employees’ keystrokes and mouse movements. The company has suspended the Model Capability Initiative, not because of workers’ understandable displeasure around being (almost) perpetually monitored or for potentially breaking privacy laws, but because it caused an internal data leak. Business Insider reported that sensitive data collected through MCI, including employees’ private conversations, performance data and transcriptions, was made inadvertently available to the entire Meta staff.
“We have carefully designed this program with privacy safeguards, and while we have no indication at this time that any data was improperly accessed by Meta employees, we’re pausing it while we investigate,” a spokesperson told BI.
Despite this official line and previous statements that employees’ collected data would be “tightly controlled,” it appears Meta wasn’t quite as on top of security as it claimed. This marks the latest in a series of AI-related cybersecurity incidents for the company. Meta reps issued a similar response in March after an agentic AI took unprompted action that also dominoed into a security breach. And earlier this month, the company had to react after hackers exploited its AI customer service chatbot to hijack Instagram accounts.