Meta to Track Employee Computer Activity to Train AI Agents, Report Says

Company Aims to Improve AI Interaction With Everyday Digital Tasks

Meta is preparing to monitor how its U.S.-based employees use computers — including mouse movements, clicks, and keyboard input — as part of a broader effort to train more capable artificial intelligence agents, according to a report from Reuters.

The initiative reflects a growing challenge across the tech industry: developing high-quality, real-world data that can teach AI systems how people actually interact with software, from navigating menus to completing routine digital tasks.

Internal Program Targets Workplace Activity

According to internal memos cited in the Reuters report, the program is being developed by Meta’s Superintelligence Labs under what is called the “Model Capability Initiative.”

The tracking software will operate within designated work-related applications and websites. It will also periodically capture screenshots to provide context, helping AI systems better understand user behavior.

“This is where all Meta employees can help our models get better simply by doing their daily work,” the memo reportedly states.

A spokesperson for Meta confirmed that the collected data is intended to improve AI agents’ ability to perform common computer-based actions. Those include navigating dropdown menus, clicking buttons, and using a mouse — tasks that remain surprisingly difficult for current AI systems.

“If we’re building agents to help people complete everyday tasks using computers, our models need real examples of how we actually use them,” Meta spokesperson Andy Stone told Reuters. He added that the data will not be used to evaluate employee performance.

Privacy and Regulatory Differences Across Regions

While the program will apply to U.S. employees, implementing similar tracking in Europe could be significantly more complicated.

Strict labor and privacy laws across the European Union limit how employers can monitor workers’ activity. Meta has already faced scrutiny in the EU over its data practices, including how user-generated content is used for AI training.

The contrast highlights broader regulatory differences between the United States and Europe, where workplace monitoring policies and data consent requirements are often more tightly enforced.

The Challenge of Training AI on Human Behavior

The move underscores a key limitation in current AI development. While the internet provides vast amounts of text, images, and video for training models, capturing high-quality data about physical or digital interactions — such as how people use a mouse or navigate software — is far more difficult.

Some companies have experimented with simulated environments, including advanced physics models and hand-tracking systems, to generate this type of data. However, these approaches can fall short of replicating real-world human behavior.

Meta’s strategy suggests that direct observation of human-computer interaction may offer a more effective path forward.

Industry Push Toward Autonomous AI Agents

Meta’s initiative comes as major U.S. tech companies — including OpenAI, Google, Anthropic, and Perplexity — race to develop AI agents capable of performing tasks on behalf of users.

These tools are designed to translate natural-language instructions into actions, such as filling out forms, browsing websites, or managing files. Early testing has shown promise, but also revealed limitations in handling complex or long-running tasks.

Improving how AI understands and executes basic computer interactions is seen as a critical step toward more reliable and autonomous systems.

Broader Workforce Implications

The report also notes that Meta has begun setting internal AI usage targets for some employees, particularly engineers and developers, as part of a broader shift toward AI-driven workflows.

At the same time, the company is reportedly planning workforce reductions of up to 10% globally beginning in May, a move that aligns with ongoing restructuring efforts across the tech sector.

Conclusion

Meta’s reported plan to track employee computer activity highlights both the rapid pace of AI development and the challenges of teaching machines to navigate everyday digital environments.

As competition intensifies among U.S. tech companies, access to realistic, high-quality training data may prove decisive in shaping the next generation of AI tools — even as it raises new questions about privacy, workplace monitoring, and the future of human-computer interaction.

John Irving

“Certified food guru. Internet maven. Bacon junkie. Tv enthusiast. Avid writer. Gamer. Beeraholic.”