Meta deploys tracking tool to train AI on employee workflows

• Meta uses tracking software to collect employee data for AI training.
• Is restructuring its workforce around AI.

Meta Platforms has begun deploying internal tracking software on computers used by US-based employees, collecting data like mouse movements and keystrokes to train artificial intelligence models, according to internal memos seen by Reuters. The move forms part of a broader effort to restructure internal workflows around AI systems designed to carry out workplace tasks.

The company has begun encouraging employees to use AI tools for tasks like coding, including cases where this may reduce short-term efficiency. It has also introduced a generalised “AI builder” job designation and reorganised some engineering roles around AI-focused development.

Meta recently formed an Applied AI engineering team to develop models capable of handling software development tasks, including building and deploying systems. The company has begun transferring selected software engineers into this team.

The AI tool, known as Model Ability Initiative (MCI), operates in work-related applications and websites. One memo, posted by a staff AI research scientist in an internal channel for Meta’s Meta SuperIntelligence Labs team, said the system also captures periodic screenshots of employee screens.

Company materials state that the data is intended to address gaps in how AI models replicate human-computer interaction. This include actions like using keyboard shortcuts, interacting with dropdown menus, and managing application states. “This is where all Meta employees can help our models get better simply by doing their daily work,” the memo said.

Meta Chief Technology Officer Andrew Bosworth said in a separate internal message that the company is increasing internal data collection as part of its “AI for Work” programme, now referred to as the Agent Transformation Accelerator. “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,” Bosworth said. He added that the aim is for agents to “automatically see where we felt the need to intervene so they can be better next time.”

Bosworth said the company is building datasets and evaluation systems based on employee workflows. He did not specify how the agents would be trained but said Meta would focus on building data and evaluation systems based on workplace interactions.

Meta spokesperson Andy Stone said the data gathered via MCI would be used only for model training and not for performance assessment. “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,” he said.

Meta said safeguards are in place to prevent the collection of sensitive information, though it did not specify what categories of data are excluded.

Workforce changes and restructuring

Meta plans to reduce its global workforce by 10%, with layoffs scheduled to begin on May 20, and has indicated further reductions may follow later in the year.

Similar workforce reductions have been reported in other US technology companies. Amazon has reduced approximately 30,000 corporate roles, representing close to 10% of its office-based workforce, while financial technology firm Block has cut nearly half of its staff.

Workplace monitoring and legal considerations

The use of monitoring software in corporate environments has historically been associated with detecting misconduct or non-work-related activity, according to Ifeoma Ajunwa, a law professor at Yale University. She said keystroke logging introduces continuous monitoring of employee activity. This level of monitoring subjects white-collar employees to surveillance previously more common among delivery drivers and gig workers, she said

Ajunwa stated US federal law does not impose specific limits on workplace surveillance, though some state regulations require that employees be informed when monitoring takes place.

Valerio De Stefano, a law professor at York University in Toronto, said similar practices would likely face restrictions under European regulations, particularly where monitoring involves behavioural data like keystrokes or screen capture.

In countries like Italy, electronic monitoring tied to productivity is prohibited, while German courts have limited keystroke tracking to specific cases involving suspected criminal activity. De Stefano said such data collection practices could conflict with the European Union’s General Data Protection Regulation, and that increased awareness of workplace surveillance can change the balance of power toward employers.

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information.

AI News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.