Hallucination, Verification, and the Sanctions Risk in Legal AI
How legal AI hallucinates, why citation verification matters under ABA Formal Opinion 512 and post-Mata sanctions, and how to evaluate a vendor verification claim before signing.
16 posts in this cluster
AI Hallucination Sanctions: Lawyer Cases and How to Avoid Them
Courts have sanctioned lawyers for citing AI-fabricated cases, from Mata v. Avianca to a $110,000 Oregon penalty, with the Fifth, Sixth, and Ninth Circuits now joining in 2026. Here is the current catalog of lawyer AI hallucination cases and how to verify AI legal research.
Damien Charlotin's Hallucination Tracker, Read Like a Risk Manager
The Charlotin AI hallucination tracker lists 1,624 legal cases as of 18 June 2026. Here is how a risk manager, not a news consumer, reads the legal AI hallucination cases tracker.
AI Citation Verification: How Legal AI Verifies Citations in 4 Layers
How legal AI verifies citations: the four layers of AI citation verification (exact match, citation resolution, meaning, cross-check), what each catches, and how to test any vendor's verifier.
How to Verify AI Legal Citations Before You File (ABA 512 Checklist)
How to verify AI legal citations before filing: a Rule-by-Rule ABA 512 checklist that catches fake cites, wrong quotes, and unsupported propositions.
Legal AI With No Hallucination: The AI That Doesn't Make Up Cases (2026)
Which legal AI does not hallucinate cases, and how to test for it. What the most accurate legal AI gets right: grounded retrieval, real opinions, citation verification, not model memory.
ABA Formal Opinion 512 (2024): Generative AI Duties for Lawyers
ABA Formal Opinion 512, issued July 29, 2024, sets lawyer ethics duties for generative artificial intelligence tools. A rule-by-rule guide to competence, confidentiality, fees, candor, and the duties most lawyers miss, plus the state opinions that followed.
Are ABA Formal Opinions Binding? What Opinion 512 Means for Your AI Workflow
Are ABA formal opinions binding? No. Here is how Opinion 512 actually carries weight, how states adopt or exceed it, and what it requires of your AI workflow.
"We Do Not Train on Your Data": How to Verify the Claim
Westlaw and every legal AI vendor says "we do not train on your data." Here is a 5-step checklist to verify the no-training claim in writing, not on a webpage.
Does OpenAI Train on Your Westlaw or LexisNexis Data?
No. Both LexisNexis and Westlaw publicly state your data is never used to train OpenAI's foundation models. Here is what the OpenAI partnership terms actually say, and how to verify it before you sign.
Legal Research With No Data Indexing or Human Review: A Confidentiality Checklist
Legal research with no data indexing or human review is a contract term, not a slogan. A four-part checklist to verify before you upload privileged material.
Harvey AI Subprocessors: Where Your Client Data Actually Flows
A deep read of Harvey's subprocessor list, what it discloses, and why the subprocessor chain (not the no-training claim) is the real privacy question.
Harvey Subprocessors and Beyond: How to Read a Legal AI Vendor's Subprocessor List
A practical, lawyer-first guide to reading Harvey subprocessors and any legal AI vendor's list, with the full current Harvey list, 12 things to look for, and 5 questions to ask before signing.
AI Case Law Search Explained: How Semantic Search Finds the Right Precedent
AI case law search is not magic. It is embeddings, vector distance, hybrid BM25, and re-ranking. Here is how the retrieval pipeline actually finds precedent.
AI Law Search Engine vs Keyword Search: Find Case Law Faster
An AI law search engine wins when you know the concept but not the words. Keyword and Boolean search win when you know exactly what you want. Here is when to use each.
How AI Legal Research Works: RAG, Grounding, and Citations
How AI legal research works under the hood: retrieval, grounding, citation verification, and the controls that stop an LLM from inventing cases.
Legal API vs Building a RAG Pipeline: Cost Comparison
Legal API vs building a RAG pipeline: real costs of court data, engineering time, and usage-based API pricing, compared.