Meta pauses an AI training program that tracks employees’ keystrokes after an internal leak Meta has paused its AI training program, the Model Capability Initiative (MCI), after a leak exposed sensitive employee data including private conversations and performance details. The incident, classified as SEV 2, has caused frustration among employees who are critical of the data’s initial lack of security. The company is investigating the breach, which follows other recent security incidents involving Meta’s AI systems.

• Meta is pausing its AI training program, the Model Capability Initiative (MCI), due to an internal data leak.
• The leak exposed sensitive employee data such as private conversations, performance data, and transcriptions.
• The incident was rated SEV 2, indicating a high severity.
• The MCI program, which uses employee keystrokes and mouse movements for AI model training, is mandatory for most staff.
• Employees expressed frustration and anger over the data leak and the initial lack of security measures.
• Meta stated they are investigating the incident and have no indication of improper access by employees but are pausing the program as a precaution.
• This incident is the latest in a series of security issues for Meta’s AI systems. Continue reading https://www.businessinsider.com/meta-ai-training-data-leak-exposed-employee-activity-across-company-2026-6