Meta pauses its internal AI training programme that monitors employees’ computer activity after an internal data leak exposes sensitive staff information across the company, multiple outlets report. The programme, described as the Model Capability Initiative (MCI), has collected data such as keystrokes, mouse movements and clicks, and in some reporting screen or activity-related information, which Meta uses to train AI systems. According to reports, sensitive employee data gathered through the tool was made accessible internally to a much broader group of staff than intended, after the leak. The Times of India and other outlets say the exposed material included private conversations, transcriptions, and performance-related data, affecting employees in the United States, with Meta also responding to security concerns raised by workers. Meta tells employees the programme is being stopped and says it is investigating the incident. The pause follows backlash and renewed scrutiny of how the monitoring tool handles employee privacy and security, as well as concerns that an earlier risk assessment about the tool’s security did not prevent the leak.