It’s a “gulag” at Meta: the AI race is turning into a nightmare for Mark Zuckerberg’s employees.

« It’s literally a gulag », « running a marathon in the middle of a hailstorm », « the madness of this company » : accounts describing a toxic atmosphere at Meta are multiplying. An employee even interrupted a meeting to deliver an insulting message to his boss.More broadly, between the “Model Capability Initiative,” which was strongly rejected by employees, and a major security vulnerability affecting Instagram accounts, Mark Zuckerberg’s company is experiencing serious turbulence.

Some employees of Meta’s AI division on the verge of a nervous breakdown

Meta recently held an open meeting with thousands of its employees, and let’s just say it did not seem to go well. According to the American media outlet Wired, which was able to listen to audio recordings, an employee interrupted the meeting, saying he was sick and tired of being « the company’s puppet ». He then simply asked those running the meeting to write to a Meta AI executive to tell him « he is a scumbag ».

This episode should be directly linked to the situation of Applied AI, the 6,500-person unit dedicated to AI development that was created last March. Between 30 and 50% of engineers from the core product, infrastructure and security teams were assigned to a section focused on data labeling and reinforcement learning from human feedback. In other words, these engineers found themselves doing thankless and repetitive work to design Meta’s AI models. An employee of the company told Wired: « It’s literally a gulag.Overnight, you no longer have any purpose in life ».

In fact, reports of a toxic atmosphere at Meta have multiplied in recent times. Andrew Bosworth, a chief technology officer, admitted to staff that the reorganization of the workforce around AI was « catastrophic » and that morale was at its lowest point in 20 years. For his part, Chris Cox, the chief product officer, pointed out « the madness of this company » and compared the current period to « running a marathon in the middle of a hailstorm ». At Meta, the race to develop artificial intelligence therefore seems to be taking place at the expense of employees’ mental health. And while morale does not seem to be in great shape, the company’s revenue is doing quite well: it reached $56.3 billion in the first quarter of 2026.

The atmosphere at Meta seems toxic

In the space of one week, Meta cut 8,000 jobs and reassigned 7,000 of its employees to new teams focused on developing artificial intelligence, including the Applied AI teams. Some employees also strongly objected to the « Model Capability Initiative » program, an internal system designed to capture mouse movements, clicks, and employees’ keystrokes in order to train Meta’s artificial intelligence models.In response to these objections, Mark Zuckerberg conceded a 30-minute daily pause with no data collection, under certain conditions. Meta has also faced a serious security scandal: it was discovered that, while chatting with Meta AI, it was possible to hack an Instagram account in just a few clicks. How can such a security flaw be explained? Some of the teams dedicated to security were reassigned to AI development.

For his part, Mark Zuckerberg admitted that the recent staff reorganizations had « caused stress », while stating: « we have made mistakes and will certainly make more ». Meta’s CEO also promised that there would be no more layoffs by the end of 2026 and increased the budgets allocated to corporate events.

👉🏻 Follow the tech news in real time: add 01net to your sources on Google, and subscribe to our WhatsApp channel.

Source: Wired