Meta says it is pausing its employee input tracking program after internal security issues exposed potentially sensitive data meant to train AI models
Employees had previously raised concerns about the initiative, which involves collecting workers' keystroke data to train AI models.
Lead Source
More
Discussion
TechSnif Coverage
Meta Internally Exposed Employee Data Used to Train AI Models
Meta internally exposed data from a worker-tracking program, including full prompts and private conversations.
Meta internally exposed data collected through its employee-tracking program — a system designed to gather worker keystroke data for training AI models. The exposed information included full prompts and private conversations.
Employees had already raised concerns about the tracking initiative before the exposure came to light. The program captures keystroke-level data from workers, feeding it into Meta's AI training pipeline.
The scope of what was exposed is notable: not just metadata or anonymized inputs, but complete prompts and private exchanges. That's a significant privacy problem even within an internal context.
Sources familiar with the matter provided the details to Wired. Meta has not publicly addressed the specifics of the exposure or the employee concerns surrounding the keystroke collection program.