‘It’s Undignified’: Hundreds of Workers Training Meta’s AI Could Be Laid Off

The Irish angle matters beyond headcount: Ireland is a favored EU jurisdiction for US tech contractors partly due to regulatory familiarity and labor law, so redundancies there may signal Meta is shifting annotation and evaluation work to lower-cost geographies rather than simply winding down a training phase.

This is largely disconnected from recent activity in our archive, as we have no prior coverage to anchor it to. But it belongs to a broader pattern visible across the industry: frontier labs treat human labelers and evaluators as variable costs to be dialed up during model development and cut once a training run matures. The tension here is that safety evaluation and red-teaming work, often done by the same contractor pools, is not purely a one-time cost. If Meta is also reducing ongoing evaluation capacity, that is a different risk profile than a post-training cleanup.

Watch whether a competing vendor (Scale AI, Surge, or similar) announces a significant new Meta contract in the next 60 days, which would confirm this is an outsourcing restructure rather than a genuine reduction in annotation spend. If no such contract surfaces, the more concerning read (reduced investment in ongoing human evaluation) gains credibility.