Meta has paused its internal Model Capability Initiative (MCI) program that tracked U.S. employees’ mouse movements, clicks, and keystrokes to train AI models. The pause follows revelations that sensitive employee data, including personal prompts, conversations, performance information, and medical details, was accessible to all staff due to insufficient safeguards and unencrypted storage. The tool, rolled out in April, sparked an internal high-priority security incident report (SEV). Meta stated it designed the program with privacy protections but is investigating while the rollout of the pause continues. This incident highlights the growing challenges large tech companies face in balancing internal AI development needs with robust data protection and employee privacy, especially when handling sensitive workplace monitoring data.
Meta has paused its internal Model Capability Initiative (MCI) program that tracked U.S. employees’ mouse movements, clicks, and keystrokes to train AI models. The pause follows revelations that sensitive employee data, including personal prompts, conversations, performance information, and medical details, was accessible to all staff due to insufficient safeguards and unencrypted storage. The tool, rolled out in April, sparked an internal high-priority security incident report (SEV). Meta stated it designed the program with privacy protections but is investigating while the rollout of the pause continues. This incident highlights the growing challenges large tech companies face in balancing internal AI development needs with robust data protection and employee privacy, especially when handling sensitive workplace monitoring data.