Meta suspends employee data-tracking program after internal backlash
Meta halted its Model Capability Initiative, a project that collected employee keystrokes and mouse movements for AI training, following a leak and a petition signed by more than 1,800 staff members.
Meta has suspended the Model Capability Initiative (MCI), a program that gathered employees' keystrokes and mouse movements to train internal AI models. The initiative, launched by the company's TBD AI unit in April, faced immediate resistance from staff who feared the collection of sensitive information and potential job displacement.
Workers organized a petition that attracted over 1,800 signatures, circulated memes likening the tracking system to an "Eye of Sauron," and displayed flyers referencing labor-rights protections. In response, Meta briefly allowed a 30-minute opt-out period in June, but activists deemed the measure insufficient.
The situation escalated in late June when an internal post revealed that data collected by MCI was accessible across the company, including private conversations. The post quoted a colleague stating they were "incensed" by the breach of promised security controls. Within hours, Meta turned off the tracking feature and announced that the program would remain suspended and, if reinstated, would be optional.
Meta did not comment on whether any of the collected data was used to train its AI models. The company’s chief technology officer, Andrew Bosworth, defended the program’s privacy safeguards earlier, but later statements indicated the data was more valuable than initially expected.
Industry observers said the episode could deter other firms from imposing mandatory employee data collection for AI development. William Fitzgerald, founder of The Worker Agency, described the outcome as a "historic win" for worker advocacy.