A Forced March into AI
Meta’s recent AI workforce changes refer to the large-scale reassignment of thousands of existing employees into AI-related roles and training units, combined with new monitoring programs and later partial reversals, which together reveal how one of the world’s biggest tech firms is trying to build advanced AI while testing the limits of employee consent and trust. Last month, Meta reassigned 7,000 employees into units such as an Applied AI task force to help train its coming AI models, a move many inside described as being “drafted” into an AI unit. This was not a subtle reorg; it was a blunt signal that AI now sits above everything else on the priority list. Meta had already laid off 10% of its staff, or 8,000 people, in May, which meant those who remained were told, in effect: stay, and now you work on AI.
From Forced Reassignment to “Undraft”: The Quiet Revolt
The Meta engineer reassignment did not land as a thrilling AI opportunity; many engineers compared the Applied AI task force work to data labeling and called the memo that followed an “undraft” on internal forums. In response, Meta sent a new memo saying it will now “defer to each individual’s choice” about staying in the AI training unit. People in that unit are promised preferential placement in other teams, thanks to broader staffing shortages. That is corporate-speak for: we pushed too far, morale is breaking, and we need a release valve. The reversal came only after the chief technology officer, Andrew Bosworth, acknowledged a “broader morale crisis,” noting during a recent internal session that morale was “probably one of the worst it’s ever been” in the company’s 20-year history. If you want a living case study in tech employee burnout, this is it.
The Surveillance Experiment That Backfired
If forced AI training programs were one line crossed, Meta’s Model Capability Initiative was another. This internal surveillance tool tracked staff keystrokes, mouse clicks, apps, and even took screenshots to collect data for AI training. More than 1,600 employees, including software engineers, research scientists, and designers, signed a petition demanding it stop repurposing their computer data. Their argument was blunt: you cannot talk about responsible AI while spying on your own workers. Then the worst-case scenario hit—private conversations, prompts, transcriptions, and performance reviews became visible to anyone in the company, according to an internal security notice and employee accounts. After that exposure, Meta paused the program indefinitely while it investigates. This is the kind of misstep that does lasting damage to trust; once employees see their own data mishandled, future promises of “privacy safeguards” ring hollow.
AI Ambition at Any Cost?
The pattern is clear: AI is the new core product, and everything—including labor and privacy—is being refitted around it. Meta is spending at least USD 135 billion (approx. RM621 billion) on AI infrastructure this year, alongside Amazon at USD 200 billion (approx. RM920 billion), Microsoft at USD 190 billion (approx. RM874 billion), and Alphabet at USD 185 billion (approx. RM851 billion). In leaked audio, Mark Zuckerberg argued it made sense to use Meta’s own employees for AI training because “the average intelligence of the people who are at this company is significantly higher than the average set of people that you can get to do tasks”. In other words, your day job is now part of the training data. Across the industry, employers are tracking whether workers are using company AI tools and how often. The message to tech workers is unmistakable: AI is not just another project; it is the default context of your job.
What This Means for Tech Workers and the Future of AI Labor
Meta’s AI workforce changes show a new social contract forming in big tech—one where AI ambition routinely outruns consideration for worker agency. The company first forced thousands of engineers into AI tasks, then backpedaled under pressure, offering opt-outs and transfers only after morale hit a historic low. It rolled out a surveillance-based AI training program, only to halt it after sensitive employee data leaked internally. According to the Electronic Frontier Foundation, this kind of intrusive monitoring is “an abuse of power” that highlights the need for laws protecting worker privacy. Tech workers should read this as a warning: the next wave of AI will not only change products; it will change how your work is tracked, reassigned, and valued. Companies that ignore consent and trust may ship AI faster, but they risk burning out the very people their models are learning from.
