Meta has temporarily suspended a program called the Model Capability Initiative after employee personal information was leaked through one of their AI training schemes. The initiative was designed to help Meta's AI systems better understand how humans use computers, and involved collecting employees' keyboard strokes, on-screen movements, and AI prompts (queries) for training purposes.
According to internal warnings, the leaked information includes complete employee AI prompts, transcriptions from audio files, performance-related data, and private conversations between employees. Since the program was first introduced in April, employees have criticized it as excessive workplace surveillance and have filed strong complaints against it.
For now, Meta is conducting detailed investigations into the incident and will systematically implement stronger data protection measures before relaunching the program, the company announced.
Ref: Fossbytes
