Module 4.3 · Topic 5
Legal Verification Standards
Bottom Line Up Front: Legal work demands the highest verification standard. Every citation must be checked, every factual claim verified, and every AI use documented. Professional responsibility rules require competent…
5.1 Citation and Authority Verification Protocols
When AI cites case law, statutes, or regulations, mandatory verification is the baseline. Courts have sanctioned attorneys for submitting unverified AI citations, including Mata v. Avianca Inc. (S.D.N.Y. 2023), establishing that submission of fabricated cites constitutes contempt and grounds for sanctions.
Critical warning: Submitting unverified AI-generated legal citations to a court is sanctionable conduct. For how courts have sanctioned attorneys for unverified AI output — including landmark cases and patterns in sanctionable conduct — see Module 7.1.
- Verify existence: Use a legal database (Westlaw, Lexis, Google Scholar) to confirm the case or statute exists.
- Verify holding or language: Read the actual authority to confirm AI's paraphrase matches. Mischaracterization of holdings is a common AI error.
- Check currency: Confirm the case is not reversed, vacated, or overruled. Confirm the statute is current law.
- Confirm relevance: Does the authority actually support the point AI made? If AI cited a tangential authority when a stronger one exists, the work is incomplete.
- Document each verification: Record which authorities you checked, when, and what you confirmed or corrected.
5.2 Fact-Checking for Legal Research and Analysis
Legal analysis rests on factual predicates: transaction dates, regulatory requirements, case holdings. Hallucinated legal facts are a primary risk.
When AI analyzes a case or statute, verify holdings and statutory language are accurate—read the actual authority. AI may misstate conditions, exceptions, or triggering facts. For background facts, verify against primary sources rather than trusting AI's paraphrase.
5.3 Maintaining Professional Standards With AI Assistance
Hold AI-assisted work to the same standards you apply to non-AI work: thorough research, accurate cites, sound analysis, and disclosure of limitations.
Professional standards checklist:
- All citations verified against authoritative sources before submission
- Factual claims checked against primary sources or reliable secondary authorities
- Analysis transparent—reasoning is clear and can be followed
- Limitations disclosed—client understands scope and any research bounds
- Contrary authorities discussed if material to the analysis
- Work product reflects your professional judgment, not uncritical acceptance of AI output
Professional responsibility warning:
ABA Formal Opinion 512 and state bar guidance establish that attorneys have a duty of competence in using AI, including understanding its limitations and verifying AI-generated work before submitting it to clients or courts. Using AI without verification is professional negligence.
5.4 Documentation and Audit Trails for AI-Assisted Work
Courts and bar authorities require disclosure of AI use in filings. Maintaining an audit trail demonstrates diligence and defends your work if questioned.
- Tool identification: Record which AI tool(s) you used—model name, version, date accessed.
- Prompts used: Document the prompts you gave. This shows thoughtful direction, not raw acceptance.
- Verification performed: List which citations were verified, facts cross-referenced, and analyses tested for soundness.
- Changes made: Note what you added, removed, or corrected before final work product.
- Limitations disclosed: If AI work is incomplete or depends on unverified assumptions, disclose this to the client.