Meta Implements Employee Computer Tracking to Train AI Agents

Meta has begun installing new tracking software on computers used by its US-based employees. The company aims to collect mouse movements, clicks, and keystrokes to train artificial intelligence models. This initiative is part of Meta’s broader plan to develop AI agents capable of performing work tasks autonomously, according to internal memos reviewed by Reuters.

Key Highlights

• Meta installs tracking software on US employee computers to collect data for AI training
• Model Capability Initiative records mouse movements, clicks, keystrokes, and occasional screen snapshots
• Meta plans to lay off 10 percent of its global workforce starting May 20
• Legal experts note US worker surveillance is largely unregulated compared to strict European privacy laws

AI Model Training and Data Collection

The tracking tool, called Model Capability Initiative (MCI), operates on work-related apps and websites. It also takes occasional snapshots of employees’ screens. According to a memo posted by a staff AI research scientist, the goal is to improve Meta’s AI models in areas where current systems struggle, such as navigating dropdown menus and using keyboard shortcuts.

The memo stated, “This is where all Meta employees can help our models get better simply by doing their daily work.” Meta has been integrating AI into its workflows and restructuring its workforce to focus on this technology. The company believes AI will make operations more efficient.

Chief Technology Officer Andrew Bosworth informed employees in a separate memo that Meta would increase internal data collection as part of its “AI for Work” efforts. This initiative has been rebranded as the Agent Transformation Accelerator (ATA). Bosworth described a future where AI agents handle most work, while employees direct, review, and help improve these agents.

While Bosworth did not detail the training process, he emphasized the importance of building data and evaluations for all types of work interactions. Meta spokesperson Andy Stone confirmed that MCI data would be used for model training only, not for employee performance reviews. Stone stated that safeguards exist to protect sensitive content but did not specify which data types are excluded.

Industry Trends and Legal Considerations

Meta’s move to automate tasks mirrors a broader trend among US technology companies. Many firms are adopting AI tools to handle complex tasks, leading to significant workforce changes. Meta plans to lay off 10% of its global workforce starting May 20, with further cuts possible later in the year. Amazon has recently reduced its corporate staff by 30,000, nearly 10% of its white-collar workforce. In February, fintech company Block cut almost half its staff.

Internally, Meta encourages employees to use AI agents for coding and other tasks, even if it temporarily slows productivity. The company has also introduced a general-purpose job title, “AI builder,” and formed an Applied AI (AAI) engineering team to enhance coding capabilities and develop AI agents for future projects. Meta began transferring experienced software engineers to AAI earlier this month.

Historically, companies have used computer logging and screenshotting to monitor employee conduct. Yale law professor Ifeoma Ajunwa noted that logging keystrokes increases real-time surveillance for white-collar workers. US federal law does not limit worker surveillance, though some states require notification. In Europe, such monitoring would likely violate privacy laws, including the General Data Protection Regulation. Some countries, like Italy and Germany, have strict restrictions on electronic employee monitoring.