Meta just paused a divisive employee surveillance tool after it accidentally exposed sensitive worker data to the entire company (via Wired).
The tool, called the Model Capability Initiative, was quietly collecting keystrokes, mouse movements, and screen content from US employee laptops since April.
A surveillance program nobody asked for
Meta rolled out MCI to train AI systems to use computer software the way humans do, with executives arguing that employees were the best possible source for that training data.
More than 1,600 employees signed an internal petition protesting the program, warning that it introduced serious security and regulatory risks. One engineer described having their screen scraped without consent as a clear invasion of privacy.
CEO Mark Zuckerberg reportedly defended the effort in a leaked recording, saying AI models learn best from watching smart people work, and that Meta’s own staff outperformed the average contractor for this purpose.
The leak that finally forced Meta’s hand
An internal security notice revealed that data across 45,000 internal database tables had been exposed company-wide, including private conversations, full prompts, transcriptions, and performance data.
Meta spokesperson Tracy Clayton said the company has no evidence that anyone improperly accessed the data, but is pausing MCI while it investigates. Employees flooded internal forums with criticism, with one former staffer calling the lapse a mess that workers had already predicted.
Meta’s CTO Andrew Bosworth acknowledged internally that the program fell short of its own privacy review standards. This is not Meta’s first AI related security stumble either, following a March incident involving an AI agent acting without permission, and a chatbot exploit that let hackers hijack Instagram accounts.
The pause may calm tensions for now, but with morale already strained by layoffs and reorganization, Meta’s AI ambitions are clearly testing its workforce’s patience.