Revolt at Meta: AI engineers forced to label data

What happened?

According to a report by The Next Web based on recordings obtained by WIRED, employees of Meta's Applied AI unit interrupted an internal broadcast to openly complain about working conditions. The engineers, considered elite, were reassigned to data labeling tasks, a repetitive job they described as 'literally the gulag.' The protest included direct insults to a senior executive, whom they called 'trash' in no uncertain terms. This incident, which occurred in May 2025, reflects a level of tension unusual even by tech industry standards, where internal complaints are usually channeled through formal channels or anonymous leaks, not by hijacking a live broadcast.

Why is it important?

This incident exposes the tension between Meta's ambition to lead in AI and its workforce management. Reassigning top engineers to manual tasks reflects a questionable strategy that can damage morale and talent retention. Moreover, it casts doubt on the company's operational efficiency at a critical time in the AI race. Meta has invested billions in AI infrastructure, including acquiring tens of thousands of GPUs and developing models like Llama 3. However, the decision to move Applied AI engineers —a group working on practical applications of AI models— to data labeling tasks suggests a possible lack of planning or an urgency to improve training data quality. This practice is not new in the industry: Google and OpenAI have also used contractors for labeling, but they have rarely reassigned high-level employees to these tasks, indicating a possible human resources crisis.

Potential consequences

• Brain drain: Key engineers could leave Meta for competitors like Google, OpenAI, or AI startups. Low morale and the perception that their skills are undervalued could accelerate departures, especially in a labor market where AI experts are in high demand.
• Reputational damage: The 'gulag' label —a term evoking Soviet forced labor camps— could affect Meta's image as an employer. This may hinder future hiring and generate negative media coverage, as happened with the Facebook Papers leaks in 2021.
• Policy review: Meta may be forced to reassess its allocation of human resources in AI. The company might outsource more labeling or implement voluntary rotations, though the incident suggests the current direction is unsustainable.
• Impact on projects: The productivity of the Applied AI unit could suffer, delaying innovations in products like virtual assistants, recommendations, or content creation tools. If engineers are unhappy, code quality and development speed may decline.

What readers should know

This case is not isolated: in the tech industry, pressure for AI results has led to questionable labor practices. For example, in 2023, Amazon Web Services employees reported similar conditions in labeling tasks for Alexa. For investors, it is a red flag about Meta's internal culture, which had already been criticized for its management during the mass layoffs of 2022-2023 (over 21,000 employees). For industry professionals, it is a reminder that even the most valued companies can fail in talent management, especially when prioritizing speed over employee well-being. Moreover, this incident occurs in a context where Meta faces multiple challenges: lawsuits over child safety, multi-billion-dollar AI investments with no clear return, and fierce competition with Google and OpenAI. Meta's reaction —which has not officially commented— will be crucial in determining whether the company can contain the damage and retain its key talent.