Meta Is Tracking Employee Mouse Movement in 2026 — and Workers Are Fighting Back

Meta has installed tracking software on every employee laptop, and the company is no longer pretending the data stays internal. The program, called the Model Capability Initiative, captures mouse movement, keystrokes, clicks, and periodic screenshots across a hand-picked list of work apps and websites. Meta says the harvested behavior will train AI agents to navigate software the way humans do. Workers say it looks a lot like surveillance dressed up as research, and a one-week countdown to the May 20 layoffs has turned quiet grumbling into flyers, petitions, and internal Workplace threads that have spilled into the press.

What is Meta's mouse-tracking software actually doing?

The tool is officially called the Model Capability Initiative, or MCI, and it lives inside Meta SuperIntelligence Labs. According to internal messages first reported by CNBC and Fast Company, the software runs on every Meta-issued laptop and records four data streams while employees use a designated list of apps and websites. It captures mouse movement and clicks, keystrokes, and periodic screenshots of what the worker has on screen. The point, per a memo from the SuperIntelligence Labs team, is to give Meta's in-house AI agents real examples of how humans navigate everyday tools.

The tracked surface is wide. Internal documentation reviewed by reporters lists Google, LinkedIn, Wikipedia, GitHub, Salesforce's Slack, Atlassian's Jira and Confluence, and Meta's own properties including Threads and the Manus internal tool. Meta has emphasized that MCI only activates on this allowlist, not on every program running on the machine, and that the software does not read files or email attachments. The memo also told staff that any incidental personal information captured from the screen, such as a private message visible in the corner of an email window, "will not be learned by the model."

That carve-out has not calmed people down. Engineers replying inside the company's Workplace forum pointed out that screen captures still snapshot whatever is visible at the moment of the grab, which can include passwords typed into a login field, unreleased product roadmaps, immigration paperwork, health information, and personal messages from family members. One employee wrote that the program could "widely expose sensitive data" even if Meta intends to filter it later. Others pointed out that filtering happens downstream, after the raw stream has already been collected and stored.

The technical scope matters for one more reason. Meta is not running a productivity tool. Tools like Hubstaff or Time Doctor are sold as performance dashboards for managers. MCI is sold as a training-data pipeline for AI agents, which means the raw stream of clicks and keystrokes is going into model training corpora, not into a manager's review tab. That distinction is what Meta is leaning on in public, and it is also the part employees find most threatening, because the agents being trained are the same agents that the company has spent the last year promising will reshape its own engineering workforce.

How are Meta employees pushing back on the surveillance?

The internal response started on Workplace, Meta's enterprise Slack-equivalent, and moved into physical offices within days. Engadget, The Next Web, and Cybernews reported that flyers appeared in Meta's Menlo Park, New York, and Seattle offices in early May, with text criticizing the program as workplace surveillance and warning that workers may be training the AI agents that will replace them. Multiple employees told reporters the message threads on Workplace ran into the hundreds before being quietly moderated.

The framing inside the company has been blunt. CNBC quoted Meta employees describing the program as "dystopian." Cybernews documented internal posts arguing that MCI is "surveillance reframed as training data." A separate employee group circulated a petition asking Meta to make MCI opt-in, publish the full list of monitored apps, and commit in writing that data collected under the program will never be used for performance review or layoff selection. The petition cites the timing — one week before the May 20 cuts — as evidence that the rollout cannot be separated from workforce planning.

The legal angle is sharper than most outside observers realize. Several employees argued in internal threads that the program could run into U.S. labor law if it is read as monitoring organizing activity. The National Labor Relations Act protects "concerted activity" — workers discussing pay, conditions, or unionization with each other. A keystroke-and-screenshot pipeline that captures messages on Slack, LinkedIn, or personal Threads accounts could chill that activity even if Meta swears it never looks. The NLRB issued guidance in 2022 making clear that broad electronic monitoring without disclosure can itself be an unfair labor practice. Meta has not made MCI opt-in.

Workers have also leaned on each other. Several engineers posted survival playbooks in internal channels covering how to lock down personal accounts, how to avoid logging into health insurance portals or banking sites from the work laptop, and how to keep private conversations off Workplace and on personal devices. The unspoken assumption in those threads is that Meta's promise about "incidental personal information" is a promise about model training, not a promise about who at Meta can pull up a screenshot.

The company has not publicly named anyone behind the protests. The flyers are anonymous, the petitions are anonymous, and the Workplace posts have been screenshotted by reporters with names redacted. That anonymity itself is a tell. Meta employees, even ones with millions of dollars in unvested stock, are willing to organize, but only with their faces hidden.

How does Meta's surveillance fit the broader 2026 bossware trend?

Meta is not inventing this. The bossware market has been growing for five years, and the company is just the highest-profile employer to fold it into an AI training story. Microsoft built Productivity Score directly into Microsoft 365 in 2020 before scaling it back under privacy pressure. Hubstaff, Time Doctor, ActivTrak, and Teramind have all become standard issue at remote-friendly mid-market employers. Most of these tools record clicks, idle time, app usage, and periodic screenshots. The category got its public face moment in June 2024, when Wells Fargo fired more than a dozen wealth-management employees for "simulation of keyboard activity" — the bank had detected mouse jigglers and treated them as misconduct.

The Meta story sits at the intersection of two trends. The first is the long bossware drift, where surveillance software keeps getting cheaper, more granular, and more normal. The second is the AI training-data scramble, where every large model lab needs real human workflow data to teach agents how to use a browser. Anthropic, OpenAI, and Google DeepMind have all paid contractors and crowd workers for similar data through services like Mercor and Scale AI. Meta's twist is to skip the contractor budget and harvest the data from its own salaried workforce — workers who cannot decline without quitting.

Public sentiment has hardened against this fast. Multiple surveys of monitored employees over the last two years have found higher reported stress and lower trust in management, and the white-collar workforce now treats continuous monitoring as a serious factor when weighing whether to stay. Cisco's layoff communications earlier this year leaned heavily on productivity metrics that originated in bossware tools, a pattern other Big Tech employers are now repeating openly.

The Big Tech-specific pattern is worth naming. Meta's MCI rollout follows the Colorado AI surveillance and wage-setting controversy from earlier in May, where state regulators began scrutinizing algorithmic management at large employers. It follows Amazon's productivity-tracking scandals at fulfillment centers. It follows revelations that LanguageLine was using algorithmic management to set interpreter pay. The throughline is that the workforce-tracking tools that started in warehouses and call centers have now reached the engineers and designers and product managers who once thought of themselves as untouchable knowledge workers.

That shift matters for the labor market, because the worker backlash against forced AI automation is no longer theoretical. It is showing up as petitions, flyers, and union conversations inside the most profitable company in the U.S. tech sector.

What can workers anywhere do about workplace mouse tracking?

Three states already give workers leverage. New York's electronic monitoring disclosure law, in effect since May 7, 2022, requires every private employer in the state to give every new hire written notice of monitoring before the job starts, plus a posted notice in a conspicuous place. Connecticut's law requires prior written notice of monitoring types and a conspicuously posted notice. Delaware has a similar disclosure rule. None of these laws ban surveillance — they require employers to tell workers it is happening. That distinction matters because failure to disclose can support a wrongful termination claim or an NLRB complaint.

California is the one to watch. AB 1221, the Workplace Surveillance Tools bill, would require employers to give workers written notice at least 30 days before deploying any new surveillance tool, ban facial recognition, gait analysis, emotion detection, and neural-data collection in the workplace, and bar employers from inferring immigration status, religion, disability, or political views from monitoring data. The bill would also require human review of any termination or discipline decision that relies on monitoring data and create a $500-per-violation penalty plus a private right of action. It cleared a key fiscal hurdle in 2025 and is now moving toward a floor vote. Workers in California should track it the way California workers tracked AB 5, because the disclosure window alone would have killed a rollout like MCI in its current form.

The federal layer is thinner but still useful. The NLRB's 2022 general counsel memo on intrusive electronic monitoring is the closest thing to a national rule, and it has been cited in unfair labor practice charges filed at large employers across retail, tech, and logistics. The Federal Trade Commission has signaled interest in algorithmic management cases that touch consumer privacy. Neither is a strong enough lever to stop a deployment on its own, but both are usable in a complaint posture if an employer fails to disclose what is collected.

Practically, workers can do four things now. First, ask in writing what your employer monitors and request the disclosure notice — in New York, Connecticut, and Delaware, you are legally entitled to one. Second, treat the work laptop as a public device. Do not log into personal email, banking, health portals, or immigration sites from it, and use a personal phone for any conversation you would not want a manager to read. Third, if your employer is rolling out a new surveillance program, ask whether it is opt-in, whether the data feeds into performance review or layoff selection, and whether there is a retention limit. Fourth, watch for exit signals. A new monitoring rollout that lands right before a layoff, the way MCI did at Meta, is rarely a coincidence. If you are at a company that has just installed bossware and started talking about restructuring, the calendar is telling you something.

The Meta employees fighting MCI may not win this round. But the pushback has already broken the corporate framing that surveillance is just a training-data story, and it has shown that even six-figure engineers will paper the walls with anonymous flyers when the monitoring crosses a line. That is the part the rest of the workforce should pay attention to.

People Also Asked

Q: Is Meta legally allowed to track employee mouse movement?

A: In California, where Meta is headquartered, current law allows electronic monitoring of work-issued devices as long as the employer provides notice and the activity is on company equipment, but pending AB 1221 would require 30-day advance written notice, ban several surveillance technologies outright, and require human review of any discipline based on monitoring data. New York, Connecticut, and Delaware already require written disclosure before monitoring starts. So the rollout is legal today, but it is operating inside a tightening regulatory window, and several employees have argued the lack of disclosure could support an NLRB complaint.

Q: Which states require employers to disclose workplace monitoring?

A: New York requires every private employer to give written notice of electronic monitoring to all new hires before the job begins, plus a posted notice in a conspicuous workplace location, under a law in effect since May 7, 2022. Connecticut requires prior written notice and a posted conspicuous notice. Delaware has a similar disclosure requirement. California's AB 1221 would add a 30-day advance notice requirement and several outright bans on specific surveillance technologies if it passes the floor vote.

Q: Can my employer fire me for using a mouse jiggler?

A: Yes, in most U.S. states. Wells Fargo fired more than a dozen wealth-management employees in June 2024 for "simulation of keyboard activity," meaning the bank had detected mouse jigglers and treated the activity as misconduct. Most employers classify mouse jigglers as falsifying time records, which is grounds for termination and, in regulated industries, for a Form U5 disclosure that follows the employee to their next job. There is no federal law protecting mouse-jiggler use.

Get the inside scoop on how the companies you work for actually treat their people. The Hire Ground tracks the surveillance rollouts, layoff timing, and management decisions Big Tech does not want on the front page — so you know exactly what you are walking into before you sign the offer.