Meta tracks employee keystroke data for agentic AI model training amid privacy furor

Meta has introduced a new employee monitoring tool that tracks the keystrokes and mouse movements of the company’s U.S.-based workers to generate data for training AI models. The company hopes those models will help AI agents work more efficiently.

“For agents to understand how people actually complete everyday tasks using computers, we need to train our models on real examples,” an internal announcement, seen by Business Insider, explains.

Many employees responded by expressing discomfort or with questions about how to opt-out of the Model Capability Initiative (MCI), which is not possible.

Reuters reports that the models in question are in development for Meta’s “AI for Work” division, which is now known as Agent Transformation Accelerator (ATA).

A Meta spokesperson told the publication that privacy protections are in place and the data will only be used for the stated purpose.

Keystroke and mouse-movement tracking is used for behavioral biometrics by companies like NeuroID and TypingDNA.

A report from the U.S. Government Accountability Office (GAO) last year warned that employee surveillance tools are reshaping privacy expectations in the workplace.

Meta employees agree to the company monitoring their work devices when they are onboarded.

And the company is looking for new growth engines, as its core social media business has been fined hundreds of millions of dollars and faces potential liability above a billion, and the backlash to its smart glasses with native facial recognition threatens to kickstart another round of expensive fees and settlements for privacy violations.

If businesses will pay for Meta’s agentic AI tools, it could yet find a business line that generates less litigation.

Article Topics

AI agents  |  behavioral biometrics  |  biometrics  |  data privacy  |  keystroke biometrics  |  Meta  |  workforce