← Back to feed

The AI restructuring at Meta is not going quietly. Meta CTO Andrew Bosworth acknowledged in an internal memo on June 16 that employee morale is near its worst in company history, comparing the current climate to the Cambridge Analytica crisis — a period in 2018 when public backlash nearly cratered Facebook’s brand and regulatory exposure spiked across three continents.

The comparison is precise. Cambridge Analytica was an external trust crisis. This one is internal, and it is structural.

What Changed

Meta has executed three overlapping transformations in under 18 months: 8,000 layoffs concentrated in middle management and product roles, a forced reassignment of roughly 10% of remaining engineers into work that supports AI model training, and the deployment of keystroke and screen-capture monitoring software on US employee computers — disclosed as a tool for improving AI agents.

Each decision has a logical justification in isolation. Collectively they compose a workplace that employees, sourced by TechCrench, are describing as a “soul-crushing gulag” inside Meta’s internal AI unit.

Bosworth’s memo attempted to acknowledge the shortcomings of how the restructuring was communicated. It did not reverse any of the decisions. The company has instead increased budgets for travel, events, and office perks including snacks as a compensatory gesture.

The Hackathon Miscalculation

Zuckerberg separately announced a companywide AI hackathon planned for July, framing it as a return to Meta’s hacker-culture roots. Employees pushed back visibly. The disconnect is direct: a hackathon asks engineers to spend discretionary energy building new things while a parallel policy monitors their keystrokes to extract data for AI training. The cultural frame of the hackathon collides with the operational reality of the monitoring program.

Bosworth’s Cambridge Analytica comparison tracks because the company recovered from that crisis by changing its external behavior — adding privacy controls, settling with regulators, revamping its data practices. The current morale problem cannot be solved with a privacy pivot. The reassignments are the product strategy, not a side effect of it.

The Pattern

Meta is not alone in this position. IBM, Intuit, and Coinbase have all announced job cuts in the same period while citing AI productivity gains. The difference at Meta is scale and transparency: Bosworth put the morale diagnosis in writing and sourced it to the company’s own history of crisis points.

The underlying dynamic is now documented. Companies that restructure heavily around AI development do not just cut headcount — they repurpose existing technical talent into the training pipeline that feeds the models replacing the roles they just eliminated. Engineers who joined to build products find themselves labeling data. The gap between what was hired for and what is now required is generating measurable cultural damage.

At Meta, that damage has reached the point where the CTO is telling staff the situation resembles one of the worst periods in the company’s modern history. That is a signal with range well beyond the standard restructuring announcement.

Numbers

• 8,000 employees laid off (approximately 10% of workforce)
• ~10% of remaining engineers shifted to AI training support roles
• Keylogging and screen-capture software deployed on US employee machines
• Morale benchmark cited by Bosworth: comparable to Cambridge Analytica era (2018)
• Employee response to July hackathon announcement: negative