Meta’s 6,500-Person Applied AI Unit Faces Employee Revolt Over Forced Transfers and ‘Soul-Crushing’ Work
Engineers say they were conscripted into a months-old division to write coding puzzles for AI training under an ultimatum and keystroke surveillance.
What matters
- Meta's months-old Applied AI division, with ~6,500 staff, is described by engineers as a 'soul-crushing' and 'literal gulag.'
- Employees report being forced into the unit under a 'join or quit' ultimatum to perform monotonous coding-puzzle work for AI training.
- Dissent escalated when an employee hijacked an internal live stream to insult management, and over 1,600 staff signed a petition against keystroke-tracking software.
- Mark Zuckerberg defended using internal engineers over contractors, claiming higher intellectual potential for training AI models.
- The scale of the unit and nature of the grievances pose material risks to Meta's AI execution and organizational stability.
Meta’s 6,500-Person Applied AI Unit Faces Employee Revolt Over Forced Transfers and ‘Soul-Crushing’ Work
Engineers say they were conscripted into a months-old division to write coding puzzles for AI training under an ultimatum and keystroke surveillance.
What happened
Meta’s Applied AI division, created just months ago and now staffed by roughly 6,500 engineers and product managers, is reportedly in open revolt. According to a TechCrunch report published Friday, employees describe the unit as a “soul-crushing gulag” and a “literal gulag,” with many claiming they were transferred into the group against their will under a “join or quit” ultimatum.
The division’s primary task is to create coding problems and puzzles used to train AI models—work that employees describe as monotonous and boring. One report, citing Wired, noted that workers feel like “conscripts” rather than volunteers. During a recent internal presentation, frustration boiled over when an unidentified employee hijacked the live stream to direct insulting language at management. Separately, more than 1,600 employees have signed a petition protesting company software that tracks their every keystroke.
At an internal meeting, Mark Zuckerberg defended the decision to staff the division with Meta engineers rather than external contractors, arguing that the intellectual potential of internal staff is significantly higher and therefore more effective for training AI models. Despite that rationale, the general mood has deteriorated sharply, and the unit is said to be on the verge of revolt.
Why it matters
A 6,500-person engineering division in crisis is not merely an HR headache; it is an operational and strategic risk. Meta is spending billions of dollars on AI development, and this unit sits at the center of that push. If the allegations of forced conscription, surveillance, and morale collapse are accurate, the company faces attrition, slowed output, and lower-quality training data at a moment when competition against OpenAI, Google, and others is intensifying.
The specific grievances—mandatory transfers, keystroke tracking, and repetitive puzzle-generation work—highlight a tension rippling across Big Tech: the scramble to amass AI training data and talent is colliding with employee autonomy and workplace culture. Companies that rely on coerced internal labor rather than motivated teams often see productivity collapse at scale. How Meta resolves this could set a precedent for how other giants staff their own AI factories.
Public reaction
No strong public signal was available from Reddit or broader social discussion channels at the time of publication.
What to watch
Watch for three things in the coming weeks. First, whether Meta publicly acknowledges the petition or the live-stream incident and announces any structural or policy changes. Second, whether the reported unrest translates into measurable attrition or internal transfers out of the Applied AI division. Third, whether Meta’s AI product release cadence slows over the next two quarters, which would suggest the dysfunction is affecting output.
Sources
- TechCrunch, “Meta’s months-old AI unit is a soul-crushing gulag, say the engineers stuck inside it,” June 12, 2026. https://techcrunch.com/2026/06/12/metas-months-old-ai-unit-is-a-soul-crushing-gulag-say-the-engineers-stuck-inside-it/
- Zamin.uz, “Meta engineers compare conditions in AI division to a ‘gulag,’” June 13, 2026. https://zamin.uz/en/technology/206829-meta-engineers-compare-conditions-in-ai-division-to-a-gulag.html
Why it matters
A new report alleges that Meta’s recently formed Applied AI division, employing roughly 6,500 people, has become a toxic workplace that engineers describe as a “literal gulag.” Employees claim they were forced into the unit under a “join or quit” ultimatum to perform monotonous AI-training work, while over 1,600 have signed a petition against keystroke-tracking software and one worker hijacked an internal live stream to insult management.
Public reaction
No strong public signal was available from Reddit or broader social discussion channels at the time of publication.
What to watch
Watch for confirming reporting, product documentation, user-visible rollout details, and credible public discussion before treating this as settled.
Sources
Public reaction
No strong public signal was available from Reddit or broader social discussion channels at the time of publication.
Open questions
- Will Meta publicly respond to the petition or restructure the Applied AI division?
- Can the reported morale collapse be tied to specific product delays or quality issues?
- How will the 'join or quit' policy affect Meta's ability to retain senior engineering talent?
What to do next
Developers
Monitor Meta AI job postings and internal sentiment; if attrition rises, expect talent redistribution across the industry.
A 6,500-person unit in distress could flood the market with experienced AI engineers or freeze hiring.
Founders
Treat this as a case study in scaling culture; rapid 6,500-person build-outs require management bandwidth and consent mechanisms that match headcount growth.
Organizational dysfunction at scale often traces to consolidation speed outpacing leadership capacity and employee agency.
PMs
Track Meta AI product release cadence over the next two quarters for signs of delay tied to internal turmoil.
Large unhappy engineering units tend to ship slower or produce lower-quality outputs.
Investors
Factor execution risk into Meta AI bets until follow-up reporting confirms or denies the revolt claims and attrition rates.
Single-source reports of this magnitude warrant caution, but 6,500 employees in distress is a material risk signal.
Operators
Audit your own rapid-consolidation playbooks; ensure management layers, cultural norms, and voluntary staffing exist before scaling past hundreds.
Months-old units with thousands of staff are vulnerable to exactly the cohesion collapse described.