Mark Zuckerberg’s Meta has paused a program that tracked employees’ every keystroke, mouse click and on-screen activity after internal data was left potentially accessible to anyone inside the company — triggering a furious backlash from more than 1,600 workers.
The program, called the Model Capability Initiative (MCI), was launched just two months ago to gather data for training artificial intelligence models. Staff were told that their computer usage — including keystrokes, mouse clicks and screen content — would be recorded and repurposed, prompting a petition demanding the company stop harvesting “employee ‘computer use’ data”. The petition warned that the tool “raises serious concerns around privacy, consent, and trust in the workplace.”
“Meta pauses employee-tracking program for AI after staff data left exposed internally, sparking privacy backlash.”
Meta halted the program on Monday after realising that some of the collected data had been left exposed to anyone within the company. A Meta spokesman confirmed the pause, saying: “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 pause follows weeks of blow-back from staff, who had already been angered by extensive layoffs and a company-wide reorganisation around AI initiatives, on which Meta is spending up to $145bn (£109bn) this year alone. In an initial attempt at damage control, Meta offered workers the option to opt out of tracking for up to 30 minutes at a time — a move one current employee dismissed as “just an attempt at damage control”. Another employee said: “I’ve never seen morale here so bad.”
Internal frustration has spilled into the open. Employees have openly insulted management using explicit language in an internal meeting on the AI-driven changes, according to a report in Wired. The same publication cited an internal security notice that referred to the exposure of data tables including “full prompts and transcriptions, private conversations, people and performance data”.
Zuckerberg had defended the tracking in an internal meeting, saying AI models learn from “watching really smart people do things” and that the coding skills of Meta engineers would dramatically improve an AI model’s abilities. But a former employee who recently left the company said the direction feels like “chasing your tail” — “exhausting and depressing”.
The company now faces the task of investigating how data meant for training AI became broadly accessible, and whether worker trust can be rebuilt after a program that many felt was forced on them without consent.