Meta paused AI training on employee keystrokes after data was 'put in a place it wasn't supposed to go:' CTO (www.businessinsider.com)

Meta has paused its Model Capability Initiative, an AI training program that tracked employee keystrokes and mouse movements, following a data leak incident. CTO Andrew Bosworth explained that the training data, designed to be secure, was mistakenly moved to an incorrect internal location by a researcher, without any indication of foul play. The pause, initially prompted by privacy concerns after sensitive employee data became accessible to the entire company, reflects broader issues surrounding employee morale and backlash against the program, which had not allowed opt-outs until recently. This incident is significant for the AI/ML community as it highlights the challenges organizations face in balancing data collection for AI training with privacy and employee consent. Bosworth admitted that the initiative failed to generate sufficient data variety, which is crucial for effective model training, underscoring the need for careful program design that emphasizes data diversity over volume. The pause serves as a reminder of the importance of establishing robust safeguards in AI training practices while maintaining a respectful and transparent relationship with employees.