Legal AI Training Jobs

Grading model output against what a competent attorney would produce: contract analysis, legal research memos, case summaries, client-facing explanations. You mark errors, explain the correct analysis and often write the reference answer. Citation checking is its own task type, verifying that cases are real and support the proposition. It is the analytical core of practice without clients, courts or billables.

Published attorney rates in this feed typically run $50 to $120 an hour, with specialist and senior roles higher. That undercuts BigLaw billing and beats document review by a wide margin. Rates rise with practice-area scarcity: tax, IP and securities work posts above general legal review.

A JD is effectively the floor; an active bar admission depends on the listing. Plenty of projects accept licensed-but-not-practising attorneys, and former practitioners are a large share of the contributor pool. Jurisdiction matters more than status: a project on Delaware corporate law wants people who practised it, and the screeners check.

Three things to clear first. Your firm’s outside-work policy, since many require disclosure of paid engagements. Conflicts: you are not representing anyone, so client conflicts rarely attach, but confidentiality terms in the marketplace NDA do. And no legal advice is being given, the work is training material, so unauthorized-practice rules are not implicated. Most moonlighting attorneys clear all three in an afternoon.

Credential checks, then a reasoning screen. Mercor interviews you with an AI that pushes on how you analyze a scenario; other platforms use written assessments graded by senior reviewers. The failure mode is answering like a treatise instead of like an advisor: the screens reward issue-spotting, clear structure and committing to a conclusion. Rehearsing that interview format before the real one measurably helps.