Legal reasoning checked by working litigators.
A brief that cites a case that does not exist is worse than no brief. We commission vetted litigators to write and grade the legal reasoning AI models learn from.
Litigation, with proof.
Legal research and drafting tools have a well-documented failure: citations that look right and do not exist, or real cases cited for propositions they do not support. Litigation models also confuse procedural rules across courts, miss deadlines that depend on jurisdiction, and overstate the strength of an argument. HUMXN commissions credential-checked litigators, judicial clerks and e-discovery specialists to write arguments, verify citations, label discovery documents and grade model output against the standard a court would apply.
Every item is written to your specification or sourced from consenting sources with consent bound to it. Documents are scanned for personal information and masked, fingerprinted at intake and checked for duplicates, and label provenance shows whether a judgment came from the author, a model suggestion or an expert reviewer. A second litigator reviews each item, and deliveries carry a signed receipt and Ed25519-signed audit trail.
For litigators, this is paid, remote work that uses the skills you exercise in practice: reading authority closely, testing arguments and spotting what the other side would attack. Rates are shown before you accept each job. The work creates data for AI systems. It is not legal advice, involves no representation of any party and creates no attorney-client relationship. Your identity is not published, and approved work is paid automatically through Stripe.
Built to your specification.
Every engagement starts from a written spec and a pilot batch. These are the most common requests we source for litigation.
Citation verification
Litigators check whether each cited authority exists, says what it is cited for and is still good law, and label every error.
Argument and motion writing
Newly written motions, briefs and memos on hypothetical facts, with the authorities and reasoning a court would expect.
Legal reasoning evaluations
Issue-spotting and analysis questions with rubrics, graded model answers and jurisdiction tags.
Discovery and privilege labels
Responsiveness, privilege and issue coding on consented or synthetic-by-hand document sets, with PII masking.
Deposition and record summaries
Expert-written summaries of transcripts and records, used to train and test long-document comprehension.
Preference judgments
Litigators compare model answers for accuracy, candor about weaknesses and correct procedure, with written rationales.
01Invented authority
Fabricated or misattributed citations are the most visible legal AI failure. Expert verification labels each one precisely.
02Procedure varies by court
Rules, deadlines and standards differ across courts. Jurisdiction-tagged data keeps models from applying the wrong rule.
03Confidential case material
Litigation files are sensitive. Consent bound to each item, PII masking and drafted-for-purpose material keep sourcing clean.
Paid work in your field.
Remote and flexible, with the rate stated before you accept. Every job is reviewed by a second expert, and approved work is paid automatically.
- A JD, LLB or equivalent law degree
- Bar admission in at least one jurisdiction
- Several years of litigation, appellate or clerkship experience
- Experience in a practice area such as commercial, employment or IP litigation
- E-discovery specialists and senior paralegals for review tasks
Verify citations
Check that each authority exists, supports the proposition and remains good law, labeling every error.
Write arguments and memos
Draft motions or memos on hypothetical facts with accurate authority and clear reasoning.
Grade model legal analysis
Score model answers against a rubric for accuracy, procedure and candor about weaknesses.
Review a colleague's work
Act as second reviewer on another litigator's labels and drafts before delivery.
Litigation: asked often.
Can you build a citation-accuracy evaluation set?
Yes. Litigators write or collect model outputs containing citations, then verify each authority: whether it exists, whether it supports the proposition, and whether it remains good law. Errors are labeled by type, giving you a measurable test of hallucinated authority. Sets are scoped by jurisdiction and practice area and start with a pilot batch.
Is this practicing law?
No. The work is creating and reviewing data for AI systems. You represent no party, give no legal advice and form no attorney-client relationship. Fact patterns are hypothetical or come from consenting sources, and your analysis is used only as training or evaluation data. Each job states its purpose before you accept.
Do I need bar admission?
Most litigation jobs require a law degree and admission in at least one jurisdiction, verified with your identity. Document review and coding tasks may be open to experienced e-discovery specialists and paralegals. Each job lists the credentials and jurisdiction it needs, and you are only offered work that matches your verified profile.
How do I get paid, and how flexible is it?
Rates are set by field and task and shown before you accept. You choose jobs that fit around your caseload and work remotely within each stated deadline. Once a second expert approves your work, it is credited to your balance and paid automatically through Stripe after identity and tax verification.
How do you verify the litigators?
Every expert passes identity verification and a credential check before receiving work, including bar admission where the job requires it. Experts may also complete a short, paid assessment in their field. After that, every job they deliver is reviewed by a second expert, and label provenance records who made and confirmed each judgment.
Can we license the data exclusively?
Yes. Commissioned litigation datasets can be licensed exclusively, with scope, term and exclusivity agreed in writing before sourcing. Data is delivered as JSONL, CSV, Parquet or Croissant with a signed receipt listing every item by ID and hash, and standing orders can extend coverage to new courts and practice areas.
Litigation, done by experts.
Need expert-made data in this field, or have the expertise to make it? Start here.