Met with strong opposition from all employees, Meta’s “refining employees” plan is forced to scale back.
Meta wants to "feed" AI with employees’ daily operations, but the employees refuse to cooperate.
According to The Information, Meta has recently made several restrictive adjustments to its employee tracking tool "Model Capability Initiative" (MCI) due to widespread resistance from employees. The adjustments include: allowing employees to pause tracking for 30 minutes at a time, expanding the scope of those exempted, and changing the data collection method from logging precise keystrokes to "activity summaries".
But these adjustments are just patches, not a halt. MCI is still running, and Meta’s AI training strategy remains unchanged.
What exactly is this tool tracking
MCI was launched in April this year by an AI research scientist at Meta SuperIntelligence Labs, Meta’s model-building team, as an internal release.
Its operation is straightforward: deploy tracking software on employees’ work computers to collect mouse movement, clicking behaviors, keyboard operations in real time, and periodically capture screen content, covering work-related applications and websites.
Why do this? Meta explains that AI models still have shortcomings in simulating human-computer interaction details—such as selecting options from drop-down menus or using keyboard shortcuts—things that human employees do every day.
Zuckerberg directly explained the logic at the all-hands meeting in April: "We are now at a stage where AI models basically learn by watching smart people do things. Letting the system observe smart people doing these things is very important."
He further explained why use in-house employees rather than outsourcing data labeling companies: "The average intelligence of this company’s employees is clearly higher than the group brought in via outsourcing companies to do tasks."
Meta spokesperson Andy Stone confirmed MCI’s data collection will be used as one of the inputs for AI training and said the data will not be used for employee performance evaluations.
How employees fought back
The resistance from employees was quite direct.
Some employees habitually ignored authorization pop-ups, while others figured out ways to disable the software via device settings. Employees also reported noticeable device slowdown after installing MCI.
Resistance quickly escalated from personal to collective action. Last month, some Meta employees posted flyers at multiple U.S. offices, urging colleagues to sign an online petition demanding Zuckerberg and company management stop the data collection. According to Reuters, the petition has been signed by over 1,500 employees.
The petition organizers also issued a statement urging the company to "treat employees and contractors with greater dignity."
What concessions Meta made, and what it held firm on
Facing the pressure, Meta announced several adjustments in an internal memo:
New pause feature: Employees can pause MCI for up to 30 minutes to "look at personal matters," after which it automatically resumes.
Expanded exemptions: Previously, only sensitive roles could apply for exemption; now it’s expanded to employees handling sensitive content, remote employees concerned about bandwidth costs, and those who work outdoors and struggle to keep their laptops charged.
Enhanced privacy protection: Data collection switched from logging precise keystrokes to "activity summaries." The company also stated that, due to strict internal access controls, only "a handful of engineers" can access raw data collected by MCI.
But the core framework hasn’t changed: MCI is still running, employees remain a source of AI training data, and the automatic resume after a 30-minute pause means this isn’t an option that can be fully turned off.
The bigger picture: Employees are being drawn into the AI pipeline
MCI is just an entry point to a larger strategy at Meta.
According to Reuters, Meta CTO Andrew Bosworth described the company’s direction in another internal memo: "The future we are building is one where agents do most of the work, and our roles are to direct, review, and help them improve." This strategic project has been named "Agent Transformation Accelerator" (ATA).
Meanwhile, Meta has begun requiring employees to use AI agents for tasks such as programming, even at the expense of short-term efficiency; the company is also breaking down barriers between some job functions and implementing a general role called "AI builder."
In other words, employees are both the source of AI training data and the ones likely to be replaced by AI in the future.
Legally, where is this allowed and where is it not
Such monitoring currently has no federal restrictions in the U.S. Yale law professor Ifeoma Ajunwa points out that state laws at most require employers to broadly notify employees of monitoring. She also notes that keystroke logging exposes white-collar employees to the kind of real-time surveillance previously reserved for delivery drivers and gig-economy workers.
In Europe, it’s a different story. Valerio De Stefano, a professor of technology and comparative labor law at York University, Toronto, says monitoring like this is very likely to be considered illegal in Europe—Italy explicitly bans using electronic monitoring to track employee productivity; German courts rule that keystroke logging can only be deployed in cases of serious criminal suspicion. De Stefano also believes this practice may violate the EU’s General Data Protection Regulation (GDPR).
De Stefano also notes that increasing employer surveillance fundamentally shifts the power balance in the workplace, further tipping it in favor of the employer.
Risk disclaimerThe market has risks, and investment should be cautious. This article does not constitute personal investment advice, nor does it take into account any user’s particular investment objectives, financial situation, or needs. Users should consider whether any opinions, viewpoints, or conclusions in this article are suitable for their particular circumstances. If you invest based on this, you do so at your own risk.