Meta Tracks Employee Computer Use for AI Training
Meta now collects data on how its employees use their work computers to improve its AI agents. The company rolled out a program named Model Capability Initiative, or MCI, on devices belonging to staff in the United States. This software operates within job-related applications and sites. It logs mouse movements, clicks, keystrokes, and takes screenshots from time to time.
How the Model Capability Initiative Works
Reuters first reported on this effort. The gathered information helps train Meta's AI systems. The goal is to make these models more skilled at handling computers like people do. That includes carrying out routine office duties similar to those done by Meta workers every day. Company representatives made clear that MCI data plays no role in evaluating employee performance.
Meta spokesperson Tracy Clayton provided a statement to The Verge. She explained the reasoning behind the collection. "If we're building agents to help people complete everyday tasks using computers, our models need real examples of how people actually use them, things like mouse movements, clicking buttons, and navigating dropdown menus," Clayton said. The company introduced an internal program to record such actions in specific programs. Measures exist to shield private information. Clayton added that the material serves no other function.
Push for Agent Transformation
Stay ahead of the AI curve
The most important updates, news, and content — delivered weekly.
No spam. Unsubscribe anytime.
Meta Chief Technology Officer Andrew Bosworth sent a memo to staff on Monday. He outlined steps to ramp up data gathering inside the firm, including through MCI. This supports the Agent Transformation Accelerator, or ATA. Bosworth described the long-term plan. "The vision we are building towards is one where our agents primarily do the work and our role is to direct, review and help them improve," he wrote.
Meta, the parent of platforms like Facebook and Instagram, has invested heavily in artificial intelligence in recent years. The firm develops large language models and explores agent-based systems that can perform complex operations autonomously. Bosworth, who joined Meta in 2006, leads technical strategy. He previously headed efforts in virtual reality hardware through Reality Labs. His memo signals a shift toward greater reliance on AI for internal operations.
This initiative comes amid broader changes at Meta. Reports indicate the company may cut thousands of jobs in May. Separately, a malfunctioning AI caused a security issue within the organization. Such events highlight the challenges of advancing AI capabilities while managing risks.
Employee monitoring for AI training raises questions about privacy and data use, though Meta stresses protections and limited scope. The MCI tool focuses only on work apps and excludes personal activities. Training data from real user behaviors could accelerate progress in agent technology, enabling more natural interactions with software interfaces.
Meta's move aligns with industry trends where tech firms seek high-quality datasets to refine AI. Human demonstrations provide valuable signals for tasks like menu navigation or form filling. By using internal sources, Meta avoids external data acquisition costs and ensures relevance to its workflows.