The company paused an internal data-collection program after staff warned that monitoring keystrokes, mouse activity and screen content raised serious privacy and security risks

Meta has paused an internal employee-monitoring program designed to collect workplace computer activity for artificial intelligence training, following a backlash from staff and concerns over how sensitive information was stored and accessed inside the company.

The program, known as the Model Capability Initiative, was created to gather data from employees’ work computers, including keystrokes, mouse movements, clicks and screen content. Meta had presented the initiative as a way to improve AI systems by using real examples of how highly skilled workers interact with software and complete tasks.

But the project quickly became controversial. More than 1,600 employees signed an internal petition opposing the tool, arguing that it created privacy, consent and security risks. Staff were particularly alarmed by reports that the program could capture personal communications, performance-related information and confidential work materials.

The pressure intensified after reporting revealed that data collected through the initiative had been exposed internally because access controls were not properly restricted. According to reports, sensitive employee data was left available to a broad group inside the company, raising questions about whether Meta had implemented adequate safeguards before launching the program.

Meta said it has found no evidence that the data was improperly accessed, but confirmed that it had paused the initiative while it investigates the security issues. The company has said the program was intended to include privacy protections and was not designed as a general surveillance tool.

The episode highlights a growing tension inside the technology industry: companies racing to build advanced AI systems increasingly want access to real workplace data, while employees are pushing back against the expansion of digital monitoring. For Meta, the controversy comes as the company is investing heavily in artificial intelligence and seeking to build AI agents capable of performing complex work tasks.

The dispute also raises broader questions about consent in the workplace. Employees using company equipment may have limited control over what data is collected, but the scale and detail of AI training data can make traditional workplace monitoring feel far more intrusive. Keystrokes, screen content and activity logs can reveal not only work patterns, but also personal messages, sensitive projects and internal debates.

For Meta’s leadership, the pause is an attempt to contain the fallout while preserving the broader goal of using internal expertise to train AI models. But for many employees, the incident has become a test of whether the company can balance its AI ambitions with privacy, transparency and trust inside its own workforce.

The controversy is unlikely to end with the temporary suspension. As AI companies seek richer data to train more capable systems, the debate over employee monitoring is likely to become a defining issue for workplace privacy across the tech sector.