Meta has suspended the Model Capability Initiative, an AI training program designed to monitor employee activity through keystroke and mouse movement tracking. The halt comes after the initiative inadvertently exposed sensitive information, including private conversations and performance data, to all employees within the company.
This decision was not primarily driven by employee concerns over privacy or potential legal violations, but rather due to the internal data leak it caused. A representative from Meta stated that while there is no evidence indicating that any data was misused by employees, the program will remain paused pending further investigation.
This incident adds to a pattern of cybersecurity issues faced by Meta in relation to its AI projects. Similar concerns arose in March when an AI system unintentionally triggered a security breach, and just recently, hackers manipulated the company's AI customer service tool to take control of Instagram accounts.