Mark Zuckerberg said Meta had made mistakes after a sweeping AI workforce shift that left thousands of engineers and product managers angry, displaced and, in some cases, openly hostile to the new setup. In an internal memo on Friday, he acknowledged that the changes had caused distress and said the company would address it.
The timing matters because the frustration inside Meta had already spilled into the open this week. A livestreamed, employee-only presentation was hijacked by an expletive-laden meltdown, and a senior Meta AI executive was told to his face to tell another executive he was “a piece of sh*t.”
At the center of the uproar is Meta’s Applied AI unit, a roughly three-month-old group of about 6,500 engineers and product managers formed as the company continues to pour billions into AI while cutting jobs elsewhere. Many employees said they learned they had been moved into the group through a surprise email, and one self-described draftee later called the process “quite random.”
The unit was created because Meta’s AI models still lacked the knowledge to outperform humans at technical tasks like coding, and the company said it needed real examples of how people complete everyday computer work. In an internal announcement reviewed by Business Insider, Meta said, “For agents to understand how people actually complete everyday tasks using computers, we need to train our models on real examples.”
Zuckerberg’s own explanation for why Meta drafted employees rather than relying on outside contractors sharpened the contrast. In a leaked audio recording from an internal meeting, he said Alexandr Wang knows the data-labeling world well and argued that the average Meta employee has “significantly higher” intelligence than third-party contractors.
That logic has done little to calm the workforce inside the new unit. Employees describe being forced in with no real choice, framing the move as join or quit, and many have taken to calling themselves draftees. The work itself has become part of the resentment: employees in the group were assigned to generate puzzles and coding problems to train AI models, which one employee called “literally the gulag” and another described as “soul-crushing.”
The complaints have not stopped with the Applied AI team. More than 1,600 Meta employees company-wide reportedly signed a petition protesting a program that monitors their clicks and keystrokes for AI training data, while Chris Cox addressed the “brutal” environment on a call with employees this week. The new organization is led by Maher Saba, a 12-year veteran of Meta and a former vice president in Reality Labs, and it reports up to Meta CTO Andrew Bosworth.
Zuckerberg’s memo points to the problem Meta now has to solve: it wants to keep building the strongest AI team possible without making the company feel like a draft board. He said Meta’s north star is to be the best place for the most talent, but the gap between that line and what employees say they are living through is still wide, and the company has not said what concrete changes will come next.
