Your Social Credit Score

26 Meta workers, an AI scoring system, and the surveillance state Americans built at work

On May 20, 2026, a scientist at Meta received an email telling her that her role had been eliminated. She was on approved pregnancy leave at the time. Her water broke the following day. She gave birth the day after that.

Two other people were cut from her team that day. One of them was likewise on leave related to pregnancy. Earlier that year, the scientist had gotten a low performance rating during a period when she’d taken leave connected to the same pregnancy.

Meta knew why she wasn’t at work. They had approved the leave themselves. But according to the federal complaint filed on her behalf, the system scored her absence as poor performance anyway.

She’s one of 26 plaintiffs. They all share the same pattern.

In May, Meta notified roughly 8,000 workers that their positions would be eliminated. But according to the complaint, managers weren’t the ones who created the list of jobs to cut. “Meta did not assemble the termination list through the considered judgment of managers who knew the work.”

One plaintiff’s own manager told him, the day before notifications went out, that he had no knowledge of who would be affected. According to the complaint, the decisions were made at the vice-president level.

The complaint describes what built the list: a constellation of systems within systems. Keystroke and activity monitoring fed into AI-scored performance rankings. Internal dashboards tracked each employee’s AI tool usage against their peers. An internal chatbot called Metamate and employee-trained “second brain” agents ingested each worker’s communications and documents. All of these tools generated data that flowed into algorithmically assisted calibration processes that, according to the complaint, had “substantially supplanted” the judgment of the managers who actually knew the work.

More than 1,000 Meta employees signed a petition demanding the company stop collecting their data to train its AI models. Petition fliers went up in Meta offices. The company’s own chief technology officer acknowledged that a “tremendous number of employees” were feeling anxieties about their futures. “It’s all bad,” he said. “I’m not going to try to sugarcoat that.”

According to the complaint, this system of AI tools produced the list. The system measured continuous digital output. People on leave can’t produce digital output. And every one of the 26 people on that list had taken, requested, or been approved for protected leave, time off guaranteed by federal or state employment law, within the previous two years. The complaint alleges that Meta never corrected for this.

One of the plaintiffs, among Meta’s heaviest AI tool users, was in the top two percent of adopters by the end of 2025. He used Claude Code and Metamate in his daily work. He was selected for termination while on approved parental leave for his second child. Meta’s system couldn’t distinguish a person on leave from a person who was just absent.

These people are now suing Meta in federal court. According to Reuters, no major American company has faced this kind of challenge before. So why does it sound so familiar?

---

The Mirror

Americans have a name for a system like this. They’ve been calling it China’s social credit system for over a decade.

Western media described a sprawling AI surveillance apparatus. A system that scored and ranked every Chinese citizen on their behavior. Low scorers got punished with travel bans and public blacklists. There was no way to appeal. The Chinese social credit system, in the Western imagination, became an example of what an authoritarian, dystopian society looked like. Columnists cited it, congressional hearings invoked it.

But there was a slight problem with that framing. The version of the system that existed in American headlines was significantly overstated.

What China had was regulatory files, blacklists, court records, corporate compliance systems, and scattered local scoring experiments. It didn’t have one national AI assigning every citizen a continuously updated number based on automated surveillance. MERICS, the leading European think tank on Chinese policy, described social credit as one of China’s “least digitised” monitoring initiatives, still heavily dependent on human investigations and decisions. Stanford’s Shorenstein Center reached similar conclusions. The Diplomat noted in 2021 that Western observers consistently inflated the system’s scope and technological sophistication.

China’s actual system is real. The blacklists and travel restrictions are real. The state’s coercive power over its citizens is vast and well documented. Nothing here is an argument that China’s surveillance apparatus is benign.

The argument is narrower.

Orwellian computer systems that constantly monitor, judge, and punish, with no humans pausing to question the output. This is what Americans feared.

The plaintiffs in this lawsuit allege Meta used a monitoring system that collected data automatically. It gave employees a score, without input from their direct managers, and fed the results into a system that helped determine who kept their job and who didn’t. And according to the complaint, the final check on the machines, the vice-presidents, took the machine at its word.

Americans imagined China had a fully automated AI surveillance scoring system. China mostly doesn’t. Meta, if the plaintiffs’ account is accurate, does.

There is, however, a distinction to be made here. China’s social credit system is operated by the government. Meta’s is operated by an employer. The Chinese state can restrict your travel and put you on a public blacklist. An employer can fire you. Those aren’t equivalent in total coercive power, and claiming they are would be wrong.

Meta’s authority comes from their role as the employer. The workers can supposedly walk away if they don’t like it. But that’s the load-bearing beam in the explanation of why one system was met with incredulity while the other got called performance analytics.

---

The Creep

Workplace monitoring is nothing new. There’s been a slow march toward this kind of system for years. It started with simple timeclocks. Then it became badge swipes. Then companies started reading employee emails. These were the first layers, and they went down easy. Your badge is company property. Your inbox is on their server.

Then came keystroke loggers and screen capture, each with a justification that made it sound routine. Data security. Productivity optimization. The system accumulated slowly.

Then the offices closed.

When COVID hit, it sent millions of workers home. Before the lockdowns and social distancing, only about 30% of large employers used digital surveillance tools on their workers. By the second year of the pandemic, that number had doubled.

The justification for the ever-increasing monitoring shifted depending on the circumstance. At the office, the explanation was security. At home, it was about trust. Managers wanted to keep an eye on their homebound employees. As a result, the once gradual creep of work surveillance tools became a stampede almost overnight.

This isn’t unique to Meta. Microsoft launched its own numerical scoring tool for employees in 2020.

By the time AI-scored productivity tools arrived, the foundation had already been laid. Each step had a name that made it sound like something other than what it was. And by 2026, according to the complaint, Meta had assembled the whole machine: a system that could monitor digital activity across every channel, measure each employee’s output against their peers, and feed all of it into a score that helped determine who got fired.

The trap was closing.

The American Psychological Association’s 2024 Work in America survey measured what this kind of monitoring does to the people living inside it. Among workers who knew they were being electronically monitored, 51 percent reported feeling tense or stressed during a typical workday, compared to 38 percent of those who weren’t monitored. 47 percent felt micromanaged, versus 23 percent. And 39 percent intended to look for a new job within the year, compared to 21 percent of unmonitored workers.

A bad manager can be argued with. A dashboard just produces a number.

These tools were sold as a way to replace managerial guesswork with data. What the data shows is that employees are more stressed, more resentful, and more likely to leave. The machine built to aid managerial judgment was undermining the conditions that make good management possible.

Privacy advocates called it surveillance. Industry called it analytics.

---

People Made the Decisions

Meta says people made the decisions. “Workforce management and organizational decisions were and are made by people, not AI.”

That sounds like a reasonable defense. So let’s have a look at it.

According to the complaint, the tools that scored and sorted employees had “substantially supplanted” the manager-driven calibration process that came before them. The vice-presidents who approved the final list didn’t know the work. The managers who knew the work weren’t consulted.

So people made the decisions. They just made them from inside a system designed to make the decisions for them. The list arrived at the VP level already decided.

The slow accumulation of systems embedded in other systems blunted the human capacity to check the machine’s output and push back. By the time someone needed to ask “should we fire a scientist on approved pregnancy leave?”, the people left in the process had stopped asking.

The machine made the call. The humans carried it out.

And outside the company, the system was never recognized for what it was. Privacy advocates flagged it. Reporters covered it as employment policy. California passed regulations. But nobody called it America’s social credit system. An engineer who quits Meta over monitoring can look around. Google does it. Microsoft, Amazon, Apple, Tesla -- they all do it. By 2021, six in ten companies with remote workers were running monitoring software. Where exactly do you go?

Even if you could leave, you might not recognize what you’re leaving. Microsoft called a workplace surveillance dashboard “Productivity Score.” Meta called its monitoring program the “Model Capability Initiative.” The names are designed to mean nothing. That’s the point.

In October 2025, California adopted regulations defining automated decision systems broadly enough to include systems that make employment decisions and systems that merely help humans make them. That rule is now part of the case against Meta.

Meta says people made the decisions. California’s regulation was written for exactly that claim. If the AI system produced the list and the humans signed off, the regulation still applies. The defense works as a press statement. As a legal strategy in California, it has a problem.

---

The first terminations take effect July 22. The scientist whose water broke the day after her notification is among them. On July 17, Judge Orrick denied the workers’ request to block them, ruling they hadn’t shown irreparable harm. The layoffs proceed. But Orrick acknowledged the case raises “serious questions” and ordered Meta to explain how specific employees ended up on the list. The next hearing is August 24.

Meta’s attorney told the court there is “no evidence that AI was used to make decisions in this reduction in force.” The company calls it speculation. The complaint calls it a system of systems that scored, sorted, and selected employees for termination while they were on legally protected leave.

The outrage was always about who held the clipboard. An adversary’s surveillance is authoritarianism. The same surveillance at home is management.

---

---

Enjoyed this piece?

I do all this writing for free. If you found it helpful, thought-provoking, or just want to toss a coin to your internet philosopher, consider clicking the button below and donating $1 to support my work.