The global brand, Meta, has initiated a new program to track the way employees work—their keystrokes and number of clicks on the mouse—to train its AI. This new decision has sparked backlash within the company itself.
Employers are informed that the new system will run on all computers and internal apps of Meta to log their activities; this data will be used as training for AI technology. As per media reports, the tool, called Model Capability Initiative (MCI), logs the mouse movements, clicks, and keystrokes. It also keeps an eye on how employees check software, including the shortcuts and dropdown menus, and also takes screenshots to inform AI on the ongoing activities of the screen.
This new software will be integrated for all the US-based full-time employees and contingent workers. Employees will get a pop-up on their laptops asking for permission to be enabled. It will run on a pre-approved list of work-related apps and websites like Google Chat, Gmail, Metamate, and development tools like VS Code.
The Meta spokesperson released a statement that if we are developing to help employees do their day-to-day tasks on a laptop, our models require real examples of how people operate. There is no other purpose, and sensitive content has been protected with complete safety.
While this monitoring will be limited to only company devices, it will not track employees’ phones. The sole purpose behind the data collection is to train AI models, not to evaluate the performance.
As reported in the BBC, a Meta employee whose smallest actions are recorded to train an AI model, the team expects additional job cuts.
Meta has already cut down 2,000 employees this year, and further employees are expecting increased job losses in the near future.
This program is part of Meta’s broader push to develop an AI agent that can do office tasks like actual employees. Recently, the company has built Meta Superintelligence Labs, encompassing a team of researchers and engineers who are actively working on AI advancement.
As stated by Meta, the present AI system does not work effectively at multitasking. They are not active in doing certain activities like choosing the right option dropdown menu, keyboard shortcuts, and working through multiple windows and apps.
To resolve this issue, the company said these models should have actual examples of how humans work on computers. With this step, Meta hopes to enhance its internal AI agent by converting the daily activities of employees into training data.
While the staff members at Meta cannot opt out of this software program, their activities on the work devices have been monitored for a long time, and they are informed of that when they sign on. So, this new program is just an extension of the ongoing rule instead of a new policy shift.
This year, in January, Mark Zuckerberg said that 2026 will be the year that AI transforms how we work. He is seeing projects that used to take huge teams now can be done by a single talented professional.
