Meta pauses internal AI training tech amid data security probe

The tool, known as Model Capability Initiative (MCI), was rolled out in April and captures mouse movements, clicks, and keystrokes on U.S.-based employees' computers to train Meta's AI models. The pause came after an employee filed a high-priority security incident report over the program's exposure of employee data. According to internal documentation, the exposed data included "full prompts and transcriptions, private conversations, people & performance data, DSS sensitivity ratings (1-4)."

Reuters had previously reported in May that the program was collecting more information than initially described and storing that data in unencrypted form, raising privacy concerns among employees. A company spokesperson, Tracy Clayton, confirmed the investigation and stated, "We have carefully designed this program with privacy safeguards and while ⁠we have no indication at this time that any data was improperly accessed by Meta employees, we're ⁠pausing it while we investigate." The tool was reportedly still recording activity as of Monday afternoon, and the spokesperson noted that the pause was rolling out and would take time to fully implement.