Meta halts AI training project tracking employees after internal security lapse

3 min readNew DelhiUpdated: Jun 23, 2026 12:59 PM IST

A Meta engineer on Monday, June 22, reportedly issued a security notice internally. (Image: Meta/Magnific)

Meta’s controversial efforts to gather employees’ computer movements to train its AI systems have been put on hold for now.

The social media giant has temporarily suspended the employee tracking programme after a worker flagged an internal security issue that exposed potentially sensitive data to other Meta employees, according to a report by Wired.

Launched in April this year, the Model Compatibility Initiative (MCI) involves an internal tool that reportedly collects computer inputs of employees such as mouse movements, click locations, and keystrokes, as well as screen content. It has only been applicable for US-based employees so far.

The roll-out of the tool was met with pushback from employees concerned about privacy, security, and personal liberty. Initially, employees could not opt out from being part of the project. However, that has reportedly changed to a limited degree after Meta employees protested, with some even circulating petitions.

In response, Meta has repeatedly argued that the data-gathering initiative is necessary in order to train its AI systems and make them better at operating computer software closer to the way humans do.

The move comes amid a broader push by tech companies to secure datasets from various sources for training AI models and AI-integrated robots. In India, for example, several factory workers have been told to wear head-mounted cameras, smart glasses, or other equipment in order to capture human movements involved in routine tasks such as folding, packing, etc. However, Meta is among the first major tech companies to direct these practices inward, treating its own workforce as a potential source for AI training data.

“We have carefully designed this programme 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,” company spokesperson Tracy Clayton was quoted as saying.

Story continues below this ad

How it unfolded

A Meta engineer on Monday, June 22, reportedly issued a security notice internally. The notice stated that databases containing information gathered by the MCI tool from employees were exposed to anyone working at the company. Soon, internal forums were buzzing with several employees expressing frustration over the security flaw. In response, Meta told staff that it is pausing MCI for the time being.

Stephane Kasriel, a Meta vice president overseeing AI research, also revealed that the security issue had been initially discovered on Thursday, June 18. While Meta engineers issued a security patch within four hours, the vulnerability persisted and access to the potentially sensitive data could not be locked down.

“We will only re-enable MCI when we are confident in the effectiveness of our data protection controls,” Kasriel was quoted as saying. Meta would be sharing more about the future of MCI because it had now “gathered sufficient data to assess the long-term value of the tool,” he added.