Meta has paused an internal employee-tracking program after sensitive data from the initiative was left exposed to its own workforce, according to Wired.
The program collects workers' keystroke data to help train the company's AI models. According to Wired, the exposed information included full prompts and private conversations — material that, per the reporting, was accessible internally rather than properly restricted.
Wired reports that employees had already raised concerns about the initiative before the exposure came to light. The decision to pause it followed the company leaving potentially sensitive data from the program accessible inside Meta.
The reporting, surfaced via Techmeme and based on Wired's sourcing, frames the episode as a misstep in how Meta handled data gathered from its own staff. The keystroke collection was intended as training fuel for AI development, but the internal exposure turned a controversial data-gathering effort into a privacy and security problem affecting employees themselves.
Beyond the specific figures and quotes in Wired's account, the available details center on three points: the program tracks employee keystrokes for AI training, the underlying data was exposed internally, and Meta responded by pausing the effort.
Why it matters: As tech companies race to feed AI systems with ever more data, this case shows how that hunger can collide with the privacy of a company's own workers — and how a security lapse can force a fast retreat.