Menlo Park: Meta Platforms is reportedly implementing a system that collects detailed records of employee computer activity in an effort to develop more capable artificial intelligence agents. The initiative, known internally as MCI, is designed to help train AI systems to understand how people interact with computers and perform workplace tasks.

The project forms part of CEO Mark Zuckerberg’s broader vision of building advanced AI agents capable of handling increasingly complex workflows. However, the programme has also sparked debate over employee privacy and data collection practices.

AI training project focuses on workplace behaviour

According to reports, the system has been deployed on devices used by Meta employees in the United States.

The objective is to gather information about how workers use software, navigate applications and complete tasks on their computers. Such behavioural data could help Meta develop AI agents capable of automating a wide range of workplace activities.

Industry experts view this approach as part of a growing trend in artificial intelligence development, where companies seek real-world interaction data to improve AI performance and decision-making capabilities.

Employees raise concerns over data usage

In the weeks following its rollout, some employees reportedly expressed concerns about the amount of data being collected.

Reports indicate that the monitoring system generated significant internet traffic, with some workers claiming it consumed large portions of their monthly home internet data allowances within a short period.

Employees also raised questions about the scope of information being gathered and how that data would ultimately be used in training future AI systems.

Non-US communications may also be captured

One of the key concerns highlighted in internal discussions relates to communications involving employees outside the United States.

According to reports, Meta acknowledged that emails and direct messages sent to US-based employees could be captured by the system regardless of where the sender is located. This has raised questions about cross-border data collection and compliance with privacy regulations in different jurisdictions.

The issue has drawn attention because the project is officially focused on US employees while potentially involving communications from colleagues around the world.

Meta says privacy risks were considered

Meta spokesperson Dave Arnold reportedly stated that the monitoring software was installed only on devices used by US employees.

According to the company, the programme is intended to analyse how people interact with computers rather than monitor the specific content displayed on screens.

Meta said non-US employees were informed that the tool had been deployed on devices used by US colleagues with whom they might communicate during normal business operations.

The company further stated that privacy risks were assessed and mitigated during both the development and deployment phases and that the initiative is intended to comply with applicable laws and regulations.

Reports detail extensive data collection

According to internal analyses cited in reports, the monitoring system may have access to a wide range of computer activity data.

The information reportedly includes code modifications, device sleep and wake cycles, website visits and clipboard activity. Concerns have also been raised regarding how some of this information is stored and protected.

Employees have questioned whether such extensive data collection could eventually be used to train AI systems capable of performing tasks currently carried out by human workers.

Part of broader AI agent strategy

Meta has invested heavily in artificial intelligence over the past several years and has increasingly focused on developing AI agents capable of assisting with everyday tasks.

The company’s long-term vision involves creating AI systems that can understand workflows, navigate software interfaces and execute multi-step tasks autonomously.

Supporters of the approach argue that access to real-world workplace interactions could significantly improve the effectiveness of future AI agents. Critics, however, continue to call for stronger safeguards around employee privacy and data usage.

Conclusion

Meta’s reported employee activity monitoring programme highlights the growing tension between AI development and privacy concerns. While the company views the initiative as a way to build more capable AI agents, questions surrounding data collection, transparency and employee rights are likely to remain central to the debate as AI technologies continue to evolve.