Internal reports indicate that Meta has begun taking unprecedented steps to develop its technological systems by currently installing new tracking software on employee computers in the United States. This software is designed to capture mouse movements, clicks, and keystrokes to use as baseline data for training the company’s artificial intelligence models. This move is part of a broad initiative aimed at building AI agents capable of independently performing daily work tasks, according to internal memos reviewed by the news agency.
Model capabilities initiative and technological objectives
The new tool is named the “Model Capabilities Initiative” and is scheduled to run in the background of work-related applications and websites, while also taking intermittent screenshots of content displayed on employee screens. These details were revealed via an internal memo posted by one of the company’s AI research scientists in a channel dedicated to the “Meta Superintelligence Labs” team responsible for building and developing smart models.
According to the memo, the primary purpose of this tool is to improve the company’s AI models, specifically in areas where these models struggle to mimic how humans interact with computers, such as selecting items from drop-down menus and using keyboard shortcuts. The memo stated that this is an area where all Meta employees can help the models improve simply by doing their usual daily work.
Strategic shift in the work environment
The parent company of Facebook and Instagram is aggressively moving toward integrating artificial intelligence into its workflows and reshaping its workforce around this modern technology, justifying the shift by claiming it will make the company operate more efficiently. In a separate memo shared by Andrew Bosworth, Meta’s Chief Technology Officer, he told employees that the company would intensify internal data collection as part of AI integration efforts, which have now been renamed the “Agentic Transformation Accelerator”.
Bosworth said in his message that the company’s vision is for agents to do the primary work, while the human element’s role is limited to guiding them, reviewing their work, and helping them improve. He added that the goal is for these agents to automatically spot where human intervention is needed so they can perform the task better next time. Although Bosworth did not explicitly detail how these agents will be trained, he emphasized that the company will be rigorous in building data and evaluations for all types of employee interactions during workflow. Meanwhile, Andy Stone, a Meta spokesperson, acknowledged that the data collected by the new tool will be among the inputs used in this context.
Workforce impacts and layoff fears
Stone clarified that the collected data will not be used to evaluate employee performance or for any purpose other than training AI models, stressing that safeguards exist to protect sensitive content, though he did not elaborate on the types of data that will be excluded from the collection process. He added that if the company is building agents to help people complete everyday tasks using computers, the models need real-world examples of how actual people use them.
This push toward automating jobs previously performed by human workers reflects a broader trend among major US companies this year, particularly in the tech sector. AI tools have captured expert attention for their ability to handle complex tasks, triggering heavy sell-offs in traditional software stocks and prompting some executives to plan sweeping job cuts.
Within this framework, Meta plans to lay off 10 percent of its global workforce starting May 20, 2026, and is considering additional major cuts later. Similarly, Amazon has trimmed 30,000 employees in recent months, representing about 10 percent of its corporate staff, while fintech firm Block has laid off nearly half of its workforce.
Structural changes and privacy concerns
Internally, Meta is urging employees to use AI agents for coding tasks and other work. The company is also blurring the lines between certain jobs in favor of a new title called “AI Builder”. It has established a new applied AI engineering team aimed at improving the coding capabilities of smart models.
However, these technologies raise serious concerns. Ifeoma Ajunwa, a law professor at Yale University, says computer activity tracking and screen-capture technologies have historically been used to monitor employee misconduct, noting that the current move subjects white-collar workers to a degree of real-time surveillance previously experienced only by delivery workers and independent contractors.
In contrast, Valerio De Stefano, a law professor at York University, believes European law is likely to ban such surveillance, explaining that using electronic monitoring to track employee productivity is strictly restricted in countries like Italy and Germany and could be considered a violation of Europe’s General Data Protection Regulation (GDPR).
Frequently Asked Questions
What is the goal of Meta’s new tracking tool?
The tool, called the “Model Capabilities Initiative,” aims to record employee mouse movements and keystrokes to gather realistic data that helps train AI models to independently perform daily work tasks.
Will the tracking data be used to evaluate employee performance at Meta?
No, a company spokesperson confirmed that the collected data will not be used to evaluate employee performance or for any purpose other than developing and training AI models, with safeguards in place to protect sensitive data.
How do legal experts view this move from a privacy perspective?
Experts warn that this step imposes unprecedented oversight on corporate employees, noting that such practices could be considered illegal or a direct violation of data protection regulations in several European countries.
