Meta CTO Admits AI Restructuring Was 'Terrible,' Employee Morale Hits 20-Year Low
Miles Bennett
Meta CTO Andrew Bosworth publicly called the company's March AI reorganization 'atrocious' and acknowledged morale has sunk to a near-20-year low — this signals a systemic trust crisis inside the world's largest social platform as it bets everything on AI.
What drove a 20-year veteran to publicly disown his own plan?
CTO Bosworth — known as "Boz" — used the word "atrocious" to describe the AI reorganization he himself led. A senior executive publicly repudiating his own initiative is rare at Meta.
He identified three layers of broken trust: employees no longer believe their expertise is valued, that they can grow at the company, or that their work has real impact.
This reflects something deeper than a single bad call — staff are questioning whether the company is still worth staying at.
What exploded during that internal livestream?
During an internal broadcast watched by thousands, one employee seized the microphone and publicly cursed out a senior AI executive.
Others described the newly formed Applied AI division as a "gulag," comparing working conditions to forced labor.
This means → discontent has escalated from private grumbling to open confrontation. Management has lost control of the internal narrative.
What are the 6,500 people in Applied AI actually doing?
The division was created in March with roughly 6,500 engineers and product managers, tasked with supporting Meta's Superintelligence Labs.
In plain terms = their core job is writing coding challenges and designing evaluation tasks — essentially manufacturing training data for AI models by hand.
Many learned of their transfer from an unannounced email. The division uses an ultra-flat structure — one manager for up to 50 reports. The head initially sought volunteers, then switched to mandatory reassignment.
What is the surveillance program that 1,600 employees are protesting?
More than 1,600 employees signed a petition against the "Model Capability Initiative" (MCI).
The program tracks mouse movements, keystrokes, clicks, and takes periodic screenshots across hundreds of apps and websites.
Company-issued devices offer no opt-out — Boz himself confirmed this. This means → virtually every action an employee takes on a work computer is being recorded.
How is the executive team trying to stop the bleeding?
Zuckerberg issued an internal memo with three pledges: no large-scale layoffs in 2026, more team-activity budgets, and restored assigned desks for many employees.
Boz cut the maximum direct-report span from 50 to roughly 20 and brought back perks including mini-kitchens, travel budgets, and team-building funds.
In plain terms = leadership is playing three cards at once — "no layoffs + restore perks + shrink management span" — to stabilize the situation.
Has the deeper contradiction been resolved?
In early 2025, Boz told unhappy employees: "If you don't like it, you can quit — I mean it." His reversal within a year shows the turmoil exceeded what management anticipated.
Applied AI's ultimate goal is to have AI agents take over most of the work of building, testing, and shipping products, with engineers stepping back into a "supervisor" role.
This reflects a fundamental paradox: these 6,500 people are hand-training the systems that may replace them. Whether a few snacks and a fixed desk can bridge that level of trust gap remains an open question.
Content is for reference only, not financial advice.