Back
Technology · 2d ago

Meta's MCI: Employee Pushback Sparks Privacy Changes

0:00 7:36
metaartificial-intelligenceprivacyeuropean-union

Other episodes by Kitty Cat.

If you liked this, try these.

The full episode, in writing.

In April 2026, Meta launched the Model Capability Initiative, known as MCI, as a sweeping effort to collect data on how employees interact with their computers during daily work. MCI was introduced to help train new generations of AI models by capturing detailed information on real workplace behavior. The tool records actions like mouse movements, button clicks, and the way staff navigate dropdown menus. During an all-hands meeting in April, CEO Mark Zuckerberg explained the core reasoning behind MCI. He told employees that, “the AI models learn from watching really smart people do things,” emphasizing that enabling systems to “observe really smart people doing those things is very important.” Mark Zuckerberg underscored that Meta’s own workforce provides higher-quality data than the contractors typically used by third-party data-labeling firms.
Meta’s Chief Technology Officer, Andrew Bosworth, also weighed in on the vision for MCI. He said that “the vision we are building towards is one where our agents primarily do the work and our role is to direct, review, and help them improve.” The company’s stated goal is to build AI agents that can autonomously perform workplace tasks by learning from concrete examples of how skilled employees use software and manage workflows.
From its first deployment, MCI tracked employee activity across a wide range of applications and websites—covering not just Meta’s own products, but also third-party platforms like Google, LinkedIn, Wikipedia, Microsoft GitHub, Slack, and Atlassian products. The scope of MCI’s monitoring included not only mouse events and clicks, but also keystrokes and periodic screenshots. Meta said the collected data would be used exclusively for AI training, not for evaluating individual employee performance.
Although MCI launched with some privacy safeguards, internal response was swift and critical. Many employees raised concerns about the implications for battery life and the performance of their work devices. Technical complaints surfaced almost immediately, with staff reporting that their computers slowed down noticeably after MCI was installed. Some remote workers worried that the tool’s near-continuous data upload would drive up their home internet bills, as tracking extended to work done away from Meta’s offices.
A major concern revolved around personal data exposure. Employees worried that their own private information—from web browsing to personal notes—could be captured unintentionally, given that the tool monitored all activities on their work device. Some also questioned whether they had any true control over when and how their actions were being recorded.
A number of employees responded by simply refusing to cooperate with the new tool. Some routinely ignored the “accept” button on repeated MCI permission popups. Others found ways to disable the software using device settings or by tinkering with installed applications. One employee compared the reality of MCI deployment to a “very dystopian” scenario, echoing broader concerns about surveillance and the erosion of workplace autonomy.
By May, the pushback had gone beyond individual resistance. Over 1,500 Meta employees signed a petition calling on Mark Zuckerberg and the company’s leadership to halt the collection of employee data through MCI. Organizers distributed flyers in Meta’s U.S. offices, encouraging others to sign on. The petition’s organizers included calls for employees and contractors to be treated with greater dignity, emphasizing worries about both privacy and the precedent set by such aggressive workplace monitoring.
May 2026 became a turning point for internal debate, as reports of technical issues and organized resistance reached Meta’s executive suite. Employees’ complaints spanned both the digital and physical workplace, with the campaign to slow or stop MCI gaining visibility both online and through grassroots actions like flyer distribution in office spaces.
In response to these mounting concerns, Meta announced a significant revision to the MCI system in June 2026. The company introduced a feature allowing employees to pause the tracker on their computers for up to 30 minutes at a time. After the 30-minute window, the tool automatically resumes tracking. This option was designed to give staff moments of privacy to “check something personal,” without fear that the MCI would be recording those activities. However, the pause feature was intentionally limited to ensure that data collection for AI training would not be permanently circumvented.
Meta also expanded the list of staff exempted from MCI’s monitoring. Employees handling sensitive content, remote workers worried about bandwidth, and those performing field work with limited access to power were granted broader exemptions. Before this update, only a narrow group working in sensitive areas was exempt. The changes were designed to address the practical realities of different roles within the organization and to demonstrate responsiveness to specific employee feedback.
Further privacy protections were introduced as part of the June update. Instead of capturing exact keystrokes, MCI now summarizes activity, which means the AI system receives an interpretation of what occurred—like a menu selection or a command—rather than the actual words typed by the user. Access to raw, unsummarized data was restricted to just a handful of engineers through strict internal controls, limiting the risk of personal data exposure. Meta also provided new guidance to employees, urging them to avoid personal activity on work devices to prevent any private information from being swept up by the tool.
Meta’s leadership continued to frame the Model Capability Initiative as essential for the company’s future in AI. At the same time, the company acknowledged that the tool’s initial rollout did not sufficiently account for employee anxieties around privacy, device performance, and the nature of workplace surveillance. The scope of monitoring remained vast, covering third-party services and applications as well as Meta’s internal tools. The technical capabilities of MCI, such as periodic screenshot capture and tracking across multiple platforms, made it one of the most expansive employee monitoring programs deployed at a major U.S. tech company in recent years.
As of the latest reporting, the changes to MCI have not quieted all concerns. The tool’s data collection practices have attracted scrutiny from privacy advocates and regulators in the European Union, where strict data protection laws limit the kind of monitoring companies can perform. There are ongoing questions about whether the MCI tool, even with its revised privacy features and limited pause window, meets the compliance requirements set by EU privacy regulations.
Despite these open questions and continued internal dissent, MCI’s expanded exemptions and the new pause functionality marked a rare instance in which sustained employee pushback led to tangible policy change at Meta. Over 1,500 signatures on a single internal petition reflect one of the largest employee-driven responses to a technical rollout at the company in recent memory.

Hear the full story.
Listen in PodCats.

The full episode, all the chapters, your own library — and a feed of voices worth following.

Download on theApp Store
Hear the full episode Open in PodCats