Meta Halts AI Training Program After Employee Data Leak

Meta has temporarily stopped one of its internal programs used for training artificial intelligence models after an internal leak revealed widespread access to sensitive employee data within the company. This incident occurs amid internal tensions at Meta, as the company's Chief Technology Officer Andrew Bosworth, known as "Boz," recently acknowledged that employee morale is nearing historic lows due to repeated layoffs and rapid expansion in AI projects.

An internal flaw exposed employee-related information, including internal conversations, job performance data, and transcripts from various recordings, according to a report by Business Insider. Meta categorized the incident as "SEV 2" in its internal security incident classification system, where level one represents the most severe incidents. A company spokesperson confirmed the occurrence and stated that investigations are ongoing to understand the full scope of the event.

The program at the center of the controversy is called the "Model Capability Initiative" (MCI). It was designed by Meta to collect employee usage data aimed at training and improving the company’s AI models. The program was installed on devices used by employees in the United States and recorded mouse movements, clicks, keyboard presses, and monitored interactions with applications and websites used during work. Internal documents revealed that the system also periodically captured screenshots of on-screen content during work hours, raising growing privacy concerns within the company.

Participation in the program was not optional for most employees, which led to internal criticism since its launch. Many workers expressed discomfort with the continuous monitoring of their digital activity. Criticism intensified after it was discovered that sensitive data was not adequately protected or access-restricted from the outset.

Internal screenshots showed employee dissatisfaction, with many questioning why such broad access to sensitive information was permitted within the company. This security challenge is not the only one Meta has faced recently. Last month, reports surfaced about a vulnerability in the company’s AI-based chatbot that allowed attackers to take control of several high-profile Instagram accounts. Among the affected were the White House account during President Barack Obama’s administration, as well as business and U.S. government official accounts.

The recent crisis highlights the increasing challenges major technology companies face in balancing AI development with protecting user and employee privacy. Although Meta affirmed that the program was designed with privacy safeguards and that there is no evidence so far of data misuse by employees, the company decided to suspend the project temporarily until investigations are complete and security measures are reassessed.