Meta’s AI initiatives face operational hurdles amid employee unrest

Meta wanted to go all-in on AI. Its employees had other ideas.

The company’s aggressive restructuring throughout 2026 has produced a cascade of internal problems that threaten to undermine the very AI ambitions driving the changes. Forced reassignments, invasive monitoring, mass layoffs, and a measurable collapse in employee morale paint a picture of a transformation that’s generating more friction than forward progress.

The draft no one volunteered for

In March 2026, Meta created an Applied AI unit composed of roughly 6,500 engineers and product managers. The catch: many of them didn’t choose to be there. Internally, reassigned workers were labeled “draftees,” a term that captures the involuntary nature of the move with uncomfortable precision.

The unit’s primary mission was generating training data for Meta’s AI models. Employees described the work as being in a “soul-crushing gulag.” Between 30% and 50% of core engineering teams were reassigned to data tasks following the restructuring, pulling experienced engineers away from the product work they’d been hired to do.

The numbers tell a revealing story about what that shift actually produced. Internal data showed a 220% increase in code changes year-over-year, which sounds impressive until you learn that user-facing features grew by just 36%. Security and reliability incidents jumped 40%. Firefighting time surged 70%.

Surveillance and the 1,600-signature petition

Meta rolled out the Model Capability Initiative, or MCI. The program monitored employee keystrokes and tracked their activities, ostensibly to measure productivity related to AI work.

Over 1,600 workers signed a petition protesting the surveillance protocols. The pushback eventually led to some limited pauses in the MCI monitoring, though the program wasn’t scrapped entirely.

Layoffs, morale, and an all-hands meltdown

In May 2026, Meta cut approximately 8,000 jobs, representing about 10% of its total workforce. The layoffs came as the company reassessed its AI plans in light of the resistance and operational difficulties bubbling up from within.

Employee sentiment, measured through internal surveys, dropped from 74% favorability to 55%.

The tension reached a symbolic peak in June 2026, when an employee disrupted a company-wide presentation with an expletive-laden outburst directed at executives.

The $14.3B bet on Scale AI

Despite the internal turbulence, Meta invested approximately $14.3 billion for a 49% stake in Scale AI, the data-labeling and AI infrastructure company.

What this means for Meta and the industry

For investors, the gap between the 220% increase in code changes and the 36% growth in user-facing features should raise questions about execution efficiency. The 40% increase in security incidents is perhaps the most operationally concerning data point. Reassigning experienced engineers away from their areas of expertise creates real vulnerabilities in systems that billions of people use daily.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.