Inside Meta, a new AI push is turning into something employees are openly calling unbearable.
A Wired report paints a tense picture of Meta’s Applied AI unit — a three-month-old organization of roughly 6,500 engineers and product managers built to accelerate the company’s AI ambitions.
But instead of excitement, insiders describe frustration, confusion, and in some cases, outright anger.
Things escalated when an employee hijacked a livestreamed internal presentation and launched an expletive-filled outburst aimed at a senior AI executive — reportedly telling attendees to call him a “piece of sh*t.” One presenter was seen covering their face in shock.
That moment, according to reports, wasn’t isolated chaos. It was a pressure valve finally breaking.
And this is where things start to get darker.
Employees say they were reassigned into the unit through surprise emails — no interviews, no choice, just a directive: join or leave.
Many now refer to themselves as “draftees.”
Table of Contents
ToggleWhy It Matters
At the center of the controversy is how Meta is building its AI future — and who is paying the emotional cost.
The Applied AI team is reportedly tasked with generating puzzles, coding problems, and training material used to improve Meta’s AI systems.
One employee described the experience bluntly:
“It’s literally the gulag.”
Another said the work felt “soul-crushing.”
The phrase is spreading internally, reflecting a deeper identity crisis inside one of Silicon Valley’s most aggressive AI pushes.
Must Read: Google Lawsuit 2026: AI Scam Network Behind Massive Fraud Surge Exposed
Key Snapshot
| Issue | Detail |
|---|---|
| Team size | ~6,500 workers |
| Structure | Up to 50 employees per manager |
| Work focus | AI training data, coding tasks |
| Hiring method | Internal reassignment (“drafted”) |
| Sentiment | Reported morale collapse |
But the bigger issue may be what happens when AI development becomes dependent on forced internal labor pipelines rather than voluntary teams.
Hidden Problem Inside Meta’s AI Push
The turmoil isn’t limited to one group.
A separate wave of frustration is spreading across Meta after 1,600+ employees reportedly signed a petition opposing a program that tracks clicks and keystrokes for AI training purposes.
Meanwhile, leadership is trying to stabilize the narrative.
Chief product officer Chris Cox reportedly addressed what employees described as a “brutal” atmosphere.
And internally, CEO Mark Zuckerberg has already acknowledged that recent changes caused “distress,” according to internal messaging reported by Wired.
But the structural decisions behind the chaos remain in place.
Internal tension drivers:
- Sudden team reassignments without opt-in
- High-pressure AI data production targets
- Surveillance concerns tied to training data collection
- Massive restructuring under aggressive AI timelines
And at the top of the Applied AI unit sits leadership tied closely to Meta’s AI overhaul strategy, including Maher Saba and oversight from CTO Andrew Bosworth.
Contrarian View: Is This Actually “Chaos” — or AI Reality?
Not everyone agrees this is dysfunction.
Some inside the tech industry argue Meta is simply doing what most companies avoid: forcing scale at speed in AI training infrastructure.
From that lens, the so-called “gulag” framing may be emotional shorthand for something more mechanical — a high-intensity data generation system required to compete in frontier AI.
And Zuckerberg’s earlier internal reasoning adds fuel to that argument. He reportedly defended using employees over contractors, suggesting Meta workers have “significantly higher intelligence” than outsourced alternatives, making them better suited for complex AI training tasks.
Supporters of this approach argue:
- AI systems need real human problem-solving data
- Speed matters more than comfort in AI competition
- Internal talent reduces outsourcing risks
But critics counter that efficiency at this scale may come at a cost that’s already visible in morale breakdowns.
That tension is now defining Meta’s AI identity crisis.
Industry Reaction
Silicon Valley observers see this as part of a broader pattern: AI expansion colliding with workforce fatigue.
One camp sees Meta as overcorrecting after years of heavy investment in failed metaverse bets. Another sees a company trying to reclaim AI leadership at any cost.
A simplified breakdown of reactions:
- AI builders: “This is normal scaling pain.”
- Ethics critics: “This is workplace coercion.”
- Employees: “This feels forced and unsustainable.”
And underneath it all, one unresolved question remains: how long can internal pressure stay contained before it affects output quality?
What Happens Next
Meta has reportedly begun acknowledging internal strain more directly, with leadership signaling that mistakes were made in rollout and structure. But no major reversal has been announced.
The Applied AI unit continues operating, continuing its rapid production of training data and internal AI development tasks.
The real uncertainty now is whether this structure stabilizes into a high-performance engine — or continues spiraling into deeper internal resistance as AI demands scale even further.
Because in Silicon Valley’s most aggressive AI race, the question is no longer just about who builds the best model.
It’s about what it takes to build it.
And whether employees will keep accepting the cost.
Editorial disclaimer:
This article is based on publicly available reporting and descriptions from the source material provided. No facts, outcomes, or quotes have been independently fabricated. Interpretations reflect analysis of reported information and may evolve as new details emerge.