Meta wants to watch you work. Every click. Each keystroke. Every scroll. Sounds scary?
The social media giant now deploys internal software that records employees’ mouse movements, keystrokes, and button clicks on company devices. Reuters reviewed internal communications confirming the rollout. The goal, according to company executives, is to feed that behavioral data into machine learning systems designed to replicate how humans operate digital tools.
For millions of white-collar workers watching from the sidelines, the move raises an uncomfortable question: Is your employer next?
What does Meta’s AI tracking system actually do?
The initiative carries an internal name: Model Capability Initiative, or MCI. Engineers built it to do more than log mouse activity. The system also captures periodic screenshots from employee screens across work apps and websites.
According to the company, Meta’s AI tracking system is a necessary groundwork for building intelligent agents — automated systems that can navigate menus, manage files, use keyboard shortcuts, and execute multi-step digital tasks without human guidance.
One internal memo captured the company’s pitch directly: “This is where all Meta employees can help our models get better simply by doing their daily work.”
In other words, employees contribute to AI development whether they opt in or not.
The bigger vision behind the data grab
Meta’s AI tracking effort connects to something much larger inside the company. Leadership now describes a future where automated agents handle the bulk of operational work, leaving human staff to supervise and refine outputs.
Meta’s chief technology officer laid out that vision plainly for staff: “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.”
That statement signals a fundamental shift in how the company defines the role of human workers. They become editors of machine output rather than primary producers.
This thinking now drives strategy at several of the world’s largest technology companies. Meta just happens to be the first to turn workforce surveillance into a formal data collection pipeline.
Layoffs run parallel to the AI expansion
The surveillance rollout arrives at a tense moment internally. Meta plans to cut roughly 10% of its global workforce starting May 20, with further reductions potentially following later in the year.
Simultaneously, the company pushes surviving employees to integrate AI tools into daily workflows, even acknowledging that productivity may dip during the transition.
Job titles have also shifted. Meta replaced several specialized designations with a broader “AI builder” label, blurring the lines between engineering, product, and design roles.
The company also launched a new Applied AI engineering team focused on autonomous agents capable of building and shipping products independently. Cut headcount on one end. Build machine capacity on the other. That calculus increasingly defines the tech sector’s approach to workforce management.
Privacy experts raise red flags
Legal analysts and labor rights advocates reacted quickly to Meta’s AI tracking news. Their concern centers on how drastically this expands workplace surveillance beyond its traditional targets.
Historically, monitoring tools tracked warehouse staff, delivery drivers, and gig workers — roles where output measurement tied directly to physical activity. Meta now applies that same logic to corporate employees sitting at desks, writing code, designing products, and making strategic decisions.
That extension troubles privacy scholars. Constant observation can alter behavior. Employees who know someone — or something — watches their every action may suppress creativity, avoid risk, and second-guess judgment calls.
U.S. federal law currently sets few barriers to workplace monitoring as long as employers notify workers in broad terms. However, the picture looks very different overseas.
European regulations, particularly the General Data Protection Regulation, create significant hurdles. Germany and Italy both enforce strict labor protections that limit or outright ban keystroke logging unless investigations demand it. Legal experts warn that Meta’s model could face serious regulatory pushback if the company attempts a global deployment.
Meta pushes back on misuse concerns
The company insists the collected data through Meta’s AI tracking plan serves one purpose only: training AI systems.
A spokesperson addressed the criticism 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 — things like mouse movements, clicking buttons, and navigating dropdown menus.”
Meta also claims safeguards exist to prevent the misuse of sensitive content, though the company declined to specify what those safeguards cover or how the company defines sensitivity.
Critics find that vagueness troubling. Once a data collection infrastructure exists, its scope tends to expand. Today, it trains AI. Tomorrow, it could inform performance reviews, restructuring decisions, or contract negotiations.
A test case that the industry will study closely
Meta’s AI tracking move arrives at a defining moment for artificial intelligence development. Companies need enormous volumes of real-world human behavior data to improve their systems. Gathering that data from willing external sources costs time and money. Gathering it from your own employees costs almost nothing.
That economic logic makes Meta’s approach attractive to competitors watching closely. If the initiative advances without significant legal or cultural backlash, similar programs will likely spread across other large firms.
If it triggers regulatory action or internal revolt, companies may need to rethink how they source the human-generated data that powers their AI ambitions.
Either way, the outcome matters well beyond Meta’s campus.
What are your thoughts on Meta’s AI tracking plan? Does your employer have the right to monitor every action you take at work to build AI systems? Please share your thoughts in the comments below.