Meta has indefinitely paused its Model Capability Initiative, an artificial intelligence training program that deployed monitoring software across the majority of its United States workforce to capture keystrokes and mouse movements. Chief Technology Officer Andrew Bosworth detailed the suspension during an interview recorded in late June, clarifying that the halt resulted from an internal data misplacement rather than an external cybersecurity breach. Bosworth confirmed that a researcher inadvertently transferred sensitive, transformed employee data to an unauthorized internal repository. The anomaly triggered an immediate operational lockdown and a comprehensive internal audit. Management emphasized that no foul play or unauthorized external access has been detected. Launched in April, the initiative aimed to aggregate high-volume behavioral data to optimize Meta machine learning architectures. The rollout immediately provoked intense internal resistance, severely degrading employee morale and drawing widespread criticism. Compounding the operational friction, Bosworth disclosed that the program yielded insufficient data variance for effective model training, instead generating repetitive samples that offered negligible utility. In response to both the technical shortcomings and workforce pushback, Meta initially rolled out expanded opt out provisions, permitting staff to deactivate monitoring tools at will. The combination of technical inefficiency, internal data mishandling, and sustained employee dissatisfaction ultimately precipitated the program indefinite suspension. Meta security and engineering teams are currently overhauling data storage protocols and access controls to mitigate future risks. The company maintains that the program operated within strict privacy boundaries and that only a restricted number of personnel held authorized access. With no timeline established for a restart, Meta has effectively abandoned its coercive employee surveillance strategy, marking a strategic pivot toward more transparent and less invasive methods for advancing artificial intelligence development.