Meta’s plans to capture its employees’ mouse movements, clicks and keystrokes for use in training its artificial intelligence models is heightening concerns about surveillance and trust
Meta is planning to collect detailed employee activity data, including keystrokes and mouse movements, to train its artificial intelligence systems, according to a report by Reuters. The initiative is aimed at improving the performance of Meta’s AI models by capturing how work is carried out in real time.
The tool, called Model Capability Initiative (MCI), will run on work-related apps and websites and be able to take snapshots of what’s on employees’ screens, according to an internal memo Reuters obtained.
The plan reflects the growing pressure on major technology firms to secure high-quality, real-world data to stay competitive in the AI race.
Read more: Meta is building an AI Mark Zuckerberg
At the same time, the approach is likely to be interpreted by employees as a form of surveillance, said Chris Neck, Professor of Management at Arizona State University.
“This is a fascinating and important issue, and from a leadership perspective it goes far beyond the technology itself,” Neck said. “What Meta is doing sends a strong signal to employees about how the organization views them.”
Even if the intent is to train AI rather than evaluate individual performance, employees may not draw that distinction, Neck said.
“In leadership research, perception matters as much as intent,” he said. “When people feel they are being closely watched, it often signals a lack of trust, even if that was not the goal.”
Research on empowerment and self leadership consistently shows that employees perform best when they experience autonomy, ownership, meaning, and trust, Neck explained.
Impact on behavior
Monitoring behavior at such a granular level, as Meta plans, risks undermining those conditions.
“It reduces perceived autonomy, raises concerns about how one’s work is being used, and can create anxiety about long term job security,” he said, particularly if employees believe their actions are helping train systems that could replicate aspects of their roles.
Reuters reported that companies pursuing advanced AI capabilities are increasingly turning to internal data sources to refine their models. That’s because access to real-world behavioral data can provide insights into how tasks are actually completed, rather than how they’re formally described.
That strategic rationale is clear, Neck said, but it comes with trade offs.
“There is a real tension here between control and empowerment,” he said. “Organizations can gain short-term efficiency and data advantages through increased monitoring, but the long-term risk is a decline in trust, psychological safety, and engagement.”
Read more: Is your AI strategy driving employees away?
And while subtle, the resulting impact on employees’ behaviors can be significant.
“When people feel monitored, they tend to become more cautious, less willing to take risks, and less innovative,” Neck said, adding that over time, this can shift a culture “from one of commitment to one of compliance.”
Ultimately, the issue extends beyond any single company or policy, Neck said.
“The leadership question is not simply whether organizations can collect this type of data, but what collecting it does to the culture,” he said. “If employees begin to feel like tools for data extraction rather than trusted contributors, the unintended consequences can outweigh the intended benefits.”
“The same data that helps train AI systems may also reduce the human engagement that organizations depend on to perform at a high level,” Neck added.
