Meta to Monitor Employee Computer Activity for AI Model Development

Meta is installing new tracking software on computers used by its employees based in the U.S. The software aims to capture mouse movements, clicks, and keystrokes. This data will be used to train Meta’s artificial intelligence models, as part of an initiative to build AI agents capable of performing work tasks autonomously.

Credit: UNSPLASH

The tool, called Model Capability Initiative (MCI), will operate on work-related applications and websites. It will also take occasional snapshots of screen content, according to an internal memo. This memo was posted by a staff AI research scientist within the company's model-building Meta SuperIntelligence Labs team.

The stated purpose is to improve the company's AI models. The models currently struggle to replicate human computer interaction, especially with tasks like choosing from dropdown menus, and using keyboard shortcuts.

An internal memo noted, "This is where all Meta employees can help our models get better simply by doing their daily work."

Meta, owner of Facebook and Instagram, is actively integrating AI into its workflows and restructuring its workforce around the technology. The company believes this will lead to more efficient operations.

Meta Chief Technology Officer Andrew Bosworth informed employees in a separate memo that internal data collection would increase. This is part of "AI for Work" efforts, now known as Agent Transformation Accelerator (ATA).

Bosworth stated, “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 added that the aim was for agents to "automatically see where we felt the need to intervene so they can be better next time.”

Bosworth did not explicitly detail the training methods for these agents. However, he affirmed that Meta would be “rigorous” in “building up data and evals for all the types of interactions we have as we go about our work.”

Meta spokesperson Andy Stone confirmed that MCI data would serve as one of the inputs for this training.

Stone specified that MCI data would not be used for performance assessments or any other purpose apart from model training. He also noted that safeguards are in place to protect "sensitive content," without further elaboration on which types of data would be excluded.

Stone explained, "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."

The move to automate functions previously performed by human staff reflects a broader trend among major U.S. companies this year. This pattern is particularly evident within the technology sector.

AI tools have garnered significant attention in Silicon Valley for their ability to handle complex tasks. These include creating applications and organising large volumes of data with limited human oversight. This has led to a selloff in traditional software company stocks and prompted some executives to plan extensive job cuts.

Meta plans to lay off 10% of its global workforce starting on May 20. The company is also considering additional large cuts later this year.

Amazon.com has similarly reduced its corporate workforce by 30,000 employees in recent months, representing nearly 10% of its white-collar staff. Additionally, the fintech company Block chopped nearly half of its staff in February.

Internally, Meta has encouraged staffers to use AI agents for coding and other tasks, even if it causes short-term delays. The company has also removed distinctions between certain job functions, favouring a new general-purpose job title called “AI builder.”

Meta created a new Applied AI (AAI) engineering team last month. This team aims to enhance the coding capabilities of Meta’s AI models and use them to develop AI agents. These agents will perform the majority of the work to build, test, and ship future products and infrastructure at Meta.

The company began transferring “strong” software engineers into AAI earlier this month.

Ifeoma Ajunwa, a law professor at Yale University, noted that computer logging and screenshotting technology have historically been used by companies. These tools typically aim to identify employee misconduct or non-work-related activities.

Ajunwa stated that the decision to log employees’ keystrokes advances data-gathering goals further. This subjects white-collar employees to a degree of real-time surveillance previously experienced only by delivery drivers and gig workers.

Ajunwa said, “On the U.S. side, federally, there is no limit on worker surveillance.” She added that state-level laws, at most, require workers to be generally informed when employers are monitoring them.

Valerio De Stefano, a law professor at York University in Toronto who studies technology and comparative labour law, believes European law would likely prohibit such monitoring.

In some European countries, like Italy, electronic monitoring for tracking employee productivity is explicitly illegal. German courts have ruled that employers can only deploy keystroke logging in exceptional circumstances, such as suspicion of a serious criminal offence.

De Stefano added that the practice would likely be considered a violation of Europe’s General Data Protection Regulation.

More broadly, De Stefano said, employee awareness of employer surveillance shifts the balance of workplace power in the employer’s favour.

---

• Meta is implementing new software to track employee computer activity, including mouse movements, clicks, and keystrokes.

• This data will be used to train artificial intelligence models, with the goal of developing AI agents that can perform work tasks autonomously.

• Meta Chief Technology Officer Andrew Bosworth stated the vision is for AI agents to "primarily do the work," with employees directing and improving them.

Source: REUTERS