Meta’s AI Race Is Turning the Workplace Into Training Data
The Register reported May 22 that leaked audio published by More Perfect Union appears to capture Mark Zuckerberg defending Meta’s plan to monitor employee activity as part of its AI push. According to the report, Zuckerberg said internal work could teach models how strong engineers use computers, including coding and other software tasks. The system reportedly involves collecting keystrokes, mouse clicks and screenshots through a Model Capability Initiative. Meta did not confirm the audio to The Register, but the company had previously acknowledged plans to monitor employees to train AI. The report also notes the tension with recent Meta job cuts and forced reassignment of some workers toward AI roles. The core issue is not just workplace privacy. It is whether elite platform companies can convert employee workflows into proprietary model fuel, then use that model capability to reshape the same labor market that produced it.
The story is easy to file under “tech company does creepy tech-company thing.” That misses the deeper point. The AI race is not just about models sitting in a lab. It is about who gets to observe the highest-value work, compress it into software, and own the resulting operating leverage.
If the report is accurate, Meta is telling employees that their daily work is not merely labor. It is training material. The keystrokes, screenshots, mouse movements and coding habits are being treated as a strategic asset because the company believes that watching capable people use computers can teach the next generation of AI systems to do the same. That is an honest description of where the industry is going, and it should make workers, regulators and investors pay attention.
The first-order debate will be privacy. Did employees consent? Is the data anonymized? Can the system see sensitive material? Those questions matter. But the second-order debate is bigger: if a platform can legally turn internal work into model behavior, then the employment relationship changes. A worker is no longer just paid to produce output. The worker is also used to create the machine that may later reduce the bargaining power of similar workers.
That does not mean every AI tool is theft or every efficiency gain is immoral. It means the incentives are lopsided. The company owns the infrastructure, the machines, the datasets, the model weights and the product channel. The employee contributes tacit knowledge: the little decisions that do not fit neatly in a manual. Once that tacit knowledge is captured at scale, the platform decides how much of the benefit flows back to labor and how much accrues to shareholders.
Meta’s defense, as described in the report, is that it needs this advantage to win the most competitive technology race in history. That is exactly the language every dominant firm uses when it wants today’s boundary to become tomorrow’s normal. We are racing China. We are racing rivals. We cannot slow down. We cannot let compliance, consent or old categories of worker protection get in the way.
The public should be skeptical of that framing. Competition can be real and still be used as a shield. National rivalry can matter and still become a permission slip for private surveillance. The issue is not whether America should have strong AI companies. It is whether the country is going to let a handful of platforms define the rules of the workplace while they build tools that can replace, rank or discipline the same workers they are observing.
This is where the politics gets uncomfortable. Traditional labor law was built around wages, hours, safety and discrimination. The AI workplace adds a new layer: behavioral capture. The most valuable thing a company may collect from an employee is not only time. It is a high-resolution map of how expertise works. If that map becomes model property, then labor policy has to catch up.
The lesson for readers is simple. Whenever a company says “we are only using this to improve AI,” ask who owns the improvement, who carries the risk, and who gets the leverage afterward. The answer will tell you more than any glossy speech about innovation.