Home / Technology
Meta staff express privacy concerns over new internal tracking tool
Mark Zuckerberg aims to bridge the generative AI gap using worker data
Meta is reportedly set to capture employee keystrokes and mouse movements on hundreds of external websites, including Google, LinkedIn, and Wikipedia. =
According to internal messages viewed by CNBC and initially reported by Reuters on Tuesday, the "Model Capability Initiative" (MCI) allows the firm to observe staff actions on work computers.
The project, led by Meta Superintelligence Labs (MSL), aims to collect data from platforms such as GitHub, Slack, and Salesforce to train its upcoming artificial intelligence models.
While a memo from MSL sought to reassure staff regarding privacy, many employees have reportedly described the project as "dystopian."
Concerns remain that sensitive information, such as passwords and personal health data, could be exposed.
A Meta spokesperson confirmed the project, stating, "If we're building agents to help people complete everyday tasks using computers, our models need real examples of how people actually use them — things like mouse movements, clicking buttons, and navigating dropdown menus."
The company maintains that safeguards are in place to protect sensitive content and that the tool will not read internal files or attachments.
This data-gathering effort is central to CEO Mark Zuckerberg’s strategy to rival competitors like OpenAI and Google.
The initiative follows a significant recruitment drive last summer, which saw the hiring of Scale AI’s Alexandr Wang to oversee model development.
Wang’s team recently unveiled the Muse Spark model, marking the first major release in the new Muse series.
Meta continues to focus on developing AI agents capable of performing complex white-collar tasks and office-related coding.
