Meta may cut over 700 employees of contractor Covalen in Ireland who trained its AI

In Ireland, more than 700 employees of Covalen — a Meta contractor responsible for content moderation and data markup for its AI systems — face the threat of layoffs. The situation stands out because these are people whose work helped the company train safety models and automate content checks.

Who this affects

Employees of Covalen's Dublin office, which provides Meta with content moderation and data services, are at risk of being laid off. According to reports based on internal documents and employee notifications, the layoffs concern more than 700 jobs out of approximately 2,000 in the local division. Around 500 people in this group are data annotators: they review AI model responses, compare them against Meta's internal guidelines, and help the system distinguish between permissible, dangerous, and illegal content.

Employees were notified of the upcoming layoffs during a brief video call on April 27, 2026, without a full Q&A session. For Covalen, this is already the second major wave of layoffs in recent months.

What their work was

These teams did not just handle routine post or complaint reviews. Their task was to test the behavior of Meta's AI systems in complex and edge-case scenarios: seeing how the model responds to questionable requests, whether it withstands attempts to bypass protective restrictions, and whether it avoids issuing forbidden instructions. In some cases, employees had to deliberately formulate provocative requests related to suicide, child sexual abuse, and other serious topics to understand exactly where the system fails and how to improve it through additional training. Essentially, the human became the example solution that the AI was then supposed to reproduce.

"Essentially, we're teaching AI how to take away our jobs," — this is how one

Covalen employee described the situation.

The workers themselves say that the problem is not just the possible loss of income, but also the price at which this work was done. One of the interviewees described it as exhausting: all day you have to think like an offender to show the system where the boundary lies. That's why this current situation is perceived as particularly painful: people are being laid off precisely at the moment when the fruits of their labor are starting to work at scale on a daily basis.

Why Meta is changing its approach

Meta announced publicly in March 2026 that in the coming years it would be more actively implementing advanced AI systems to support users and control content. At that time, the company directly stated that it planned to gradually reduce its dependence on third-party contractors and strengthen internal systems. Against this backdrop, the new round of layoffs looks not like a local measure at a single contractor, but as part of a broader restructuring that coincided with Meta's announced layoffs of approximately 10% of its workforce.

The company explains this shift as resulting from early test results: according to Meta, the new tools better catch fraud, more accurately identify violations, and are able to scale to much broader language coverage than previous review processes. This is what drives the platform to transfer more and more of these tasks from external operational channels to internal AI systems. At the same time, the company gets an argument that this is not just about cost savings, but about improved accuracy.

• AI systems help identify up to 5,000 additional fraudulent attempts per day
• complaints about fake celebrity accounts, according to the company, have decreased by more than 80%
• in certain violation categories, new systems find twice as much problematic content as previous review teams
• such tools can work with languages spoken by approximately 98% of internet users

For the business, this looks like a transition from expensive manual review to a more centralized AI infrastructure. For workers — as a signal that Meta will rely less on external teams, even if those teams helped test, markup, and improve future automation.

The CWU union is demanding negotiations over the terms of layoffs and criticizing a rule that supposedly prevents laid-off employees from joining another Meta contractor for six months.

What this means

The Covalen story illustrates one of the harshest scenarios of AI transition: companies first hire people to collect data, establish rules, and train models, and then reduce their dependence on those same specialists once automation reaches acceptable business quality. For the market, this is another signal that professions at the intersection of moderation, markup, and trust & safety will change faster than labor regulations and employment protection mechanisms can keep up.