Particle.news

Meta Faces Revolt Over Applied AI Reorganization

Leadership admits the rollout damaged employee trust and is promising manager caps, clearer role pathways, scaled‑back device monitoring, and other fixes to try to stop mass departures.

Overview

  • Meta created a roughly 6,500‑person Applied AI unit in March 2026 to shift engineers and product managers into work that generates training and evaluation data for its generative models.
  • Many reassigned employees describe the new tasks—weekly puzzles, coding problems and repetitive labeling—as menial and demoralizing, and some called the unit a “gulag.”
  • Tensions spilled into a livestreamed internal meeting this month when an attendee launched an expletive‑filled outburst at senior AI leadership, underscoring deep workplace anger.
  • More than 1,600 staff signed an internal petition opposing software that tracked keyboard and mouse activity for training data, and Meta has reportedly scaled that program back to allow brief pauses and exemptions.
  • Top executives, including CTO Andrew Bosworth and CEO Mark Zuckerberg, acknowledged a flawed rollout and pledged fixes such as capping direct reports at about 20, letting reassigned staff apply for other roles, improving communication, and boosting local team support.