Meta halts AI training program that records employee computer activity after company-wide data leak

Meta has paused its internal Model Capability Initiative after a leak made employee data widely accessible. The incident has sharpened privacy concerns and added to unease over the company's AI-driven changes.

In Short

• Meta pauses AI training program after an internal data leak
• Private chats and performance data were reportedly exposed
• The incident adds to employee concerns over privacy and AI initiatives

It seems Meta employees won't be getting relief from low morale anytime soon. Just recently, Meta Chief Technology Officer Andrew "Boz" Bosworth said employee morale at the company was near an all-time low as mass layoffs and internal AI initiatives continued to unsettle many longtime workers. Now, according to a new report, the social media giant has paused one of those AI initiatives after sensitive employee data was reportedly accessible across the company.

According to screenshots obtained by Business Insider, Meta has temporarily halted an internal AI training program after a leak exposed employees' private conversations, performance data and transcriptions. The incident was classified as a SEV 2 on Meta's internal scale of one to five, with one being the most severe.

A Meta spokesperson confirmed the incident and said the company is investigating.

"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," the spokesperson said.

The AI project that sparked privacy concerns

The programme at the centre of the controversy is called the Model Capability Initiative, or MCI. Meta installed the software on computers used by employees in the United States to collect mouse movements, clicks and keystrokes that can be used to train its artificial intelligence models.

According to an internal memo that was previously shared by a staff AI research scientist, the tool runs on work-related apps and websites and also takes occasional snapshots of content visible on employees' screens.

For most employees, participation in the programme was mandatory. That decision had already sparked criticism inside the company, with many workers reportedly uncomfortable about having their activity recorded.

The latest leak appears to have intensified those concerns. According to screenshots seen by Business Insider, some employees were frustrated that the data had not been locked down from the start and questioned why sensitive information was accessible so broadly within the company.

Another security headache for Meta

The incident is also the latest in a series of security problems for Meta. Last month, a flaw in the company's AI chatbot reportedly allowed people to hijack multiple Instagram accounts. The issue came to light after hackers allegedly gained access to several high-profile accounts, including the Obama-era White House Instagram handle, beauty brand Sephora and the account of a senior US Space Force official.

The latest leak is unlikely to ease concerns among Meta employees, many of whom are already grappling with rapid changes driven by AI.

- Ends