Most tech companies talk abstractly about building AI from human behavior. Meta just sent its employees a memo making it concrete, and personal.
The company is installing tracking software on employee computers. It will record mouse movements, keystrokes, clicks, and screenshots. And workers cannot opt out.
The message behind the move is hard to misread. Meta is telling its own workforce that their daily behavior is now training data.
What Meta is actually doing
According to a memo sent to staff and seen by Reuters, Meta is rolling out a tool called the Model Capability Initiative. The software runs on a designated list of work apps and websites and captures how employees physically interact with their computers in real time, according to Reuters.
The memo was distributed via a channel operated by Meta Superintelligence Labs. It told employees they could “do their part to help by just doing their daily work.” The initiative sits under a broader internal program previously called AI for Work, which has since been renamed the Agent Transformation Accelerator.
More Tech Stocks:
- Morgan Stanley sets jaw-dropping Micron price target after event
- Nvidia’s China chip problem isn’t what most investors think
- Quantum Computing makes $110 million move nobody saw coming
Meta confirmed the program’s existence and defended it directly. “If we’re building agents to help people complete everyday tasks using computers, our models need real examples of how people actually use them,” a company spokesperson said, citing “things like mouse movements, clicking buttons, and navigating dropdown menus,” TechCrunch reported.
Why Meta is not allowing employees to opt out of task tracking program
One detail stands out above everything else. Employees have no option to opt out of the tracking, according to OpenTools. That transforms what might otherwise be framed as a voluntary contribution into a mandatory condition of working at the company.
Meta said safeguards are in place to protect sensitive content and that the data will not be used for any purpose beyond AI training, Fortune noted. But the absence of an opt-out has already drawn concerns about consent, morale, and the long-term implications for workplace trust.
As Gizmodo framed it plainly: workers are “essentially being told they are training the systems that will replace them.”
What Meta is trying to build
The goal behind the data collection is to train AI agents capable of performing white-collar computer tasks autonomously. That includes navigating dropdown menus, using keyboard shortcuts, and completing multi-step workflows, the kind of behavior that is difficult to simulate but easy to capture from real users.
Meta has reportedly been building AI agents designed to work alongside employees, including one built specifically for Zuckerberg himself. The company is racing to ship agentic AI products amid intense competition with OpenAI and Google.
That context matters. This is not just an internal data collection exercise. It is infrastructure for a product Meta intends to bring to market. Employees are not volunteers in a research project. They are, in effect, an unpaid data workforce.
Morris/Getty Images
The Scale AI connection
The initiative is closely tied to Meta’s broader AI restructuring. Last year, Meta acquired a 49% stake in data-labeling firm Scale AI for more than $14 billion. Scale’s former CEO, Alexandr Wang, now leads Meta Superintelligence Labs, the same division that sent the Model Capability Initiative memo to staff, according to Fortune.
That organizational structure signals how seriously Meta is treating data quality and AI agent development as strategic priorities. Wang’s background in data labeling and AI infrastructure puts the right person in charge of exactly the kind of initiative the Model Capability Initiative represents.
Key details from Meta’s employee tracking program
- Tool name: Model Capability Initiative, disclosed via internal memo to Meta staff, according to TechCrunch.
- Data captured: mouse movements, clicks, keystrokes, and screenshots on designated work apps and websites, Fortune reported.
- Opt-out option: none available to employees, according to OpenTools.
- Broader program: sits under the Agent Transformation Accelerator, formerly called AI for Work, Gizmodo noted.
- Division responsible: Meta Superintelligence Labs, led by Alexandr Wang, former CEO of Scale AI, Fortune confirmed.
- Stated purpose: training AI agents to perform white-collar computer tasks autonomously, according to TechCrunch.
The regulatory risk Meta is taking on
The policy could face a serious challenge in Europe. Meta’s approach sits in direct tension with GDPR, which requires explicit consent and proportionality in how employers collect personal data. Anonymization of the data, which Meta has pointed to as a safeguard, may not be sufficient under EU standards, OpenTools noted.
It is worth remembering that this is not Meta’s first brush with data privacy regulators. The company has faced repeated enforcement actions across Europe over how it handles user data on Facebook and Instagram. Adding employee behavioral data to that record is unlikely to make those conversations easier.
That means Meta may be inviting regulatory scrutiny at exactly the moment it is trying to accelerate AI development. A clash with European privacy regulators over employee data would be costly in both time and reputation.
What it signals about where AI is headed
This story matters beyond Meta. If the company normalizes treating employee behavior as a proprietary AI training asset, other tech firms may follow. The competitive pressure to build better agents is real, and behavioral data from real workers is genuinely more useful than synthetic alternatives.
That creates a broader question about where the line sits between employer data rights and worker privacy. Meta has given its answer. It is a line the rest of the industry will now have to respond to, whether regulators push back or not.
For investors in Meta, the move reflects a company that is willing to move fast and take on friction to stay competitive in AI. For the broader market, it is a preview of the data strategies that agentic AI will demand from every company that wants to build it seriously.
Related: Bank of America resets Meta stock forecast on deal with AMD
