AI Invented Legal Authorities in a Court Order
A US appeals court reassigned a case after AI-assisted orders cited non-existent precedent. Verification rules for AI-assisted legal and high-impact work.
Short answer
AI-generated legal citations and quotes must be checked against authoritative databases and original records — asking the same tool to verify its output is not independent review. Assign accountable people, preserve sources and versions, and use AI for drafting while reserving verification for qualified staff.
Legal decisions depend on verifiable facts and authentic authorities. On 22 October 2026, a United States federal appeals court overturned an injunction and ordered the case reassigned after a lower court's orders contained serious errors, including material apparently invented by AI.
The case concerned a Mississippi education law. The Fifth Circuit Court of Appeals recorded that an earlier order had named people who were not parties, referred to statements not in the record, misquoted legislation, and relied on a supposed legal precedent that did not exist. The trial judge had acknowledged that a law clerk used an AI tool without proper checking.
The appellate ruling was not based on the mere fact that technology had been used. The judges also identified separate legal and procedural problems. That distinction matters: AI did not automatically invalidate the work, but its unverified output helped contaminate a process in which factual accuracy and public confidence are essential.
What to do
The lesson applies to law firms, tribunals, insurers, government agencies, and businesses preparing decisions that affect other people. Every legal citation generated by AI should be checked against an authoritative legal database. Every quoted sentence should be compared with the actual judgment, contract, or statute. Important factual statements must be traced to the original record.
Review cannot mean asking the same AI tool whether its answer is correct. A persuasive confirmation may simply repeat the first mistake. Someone qualified must perform an independent check and record that it occurred. An organisation should also make clear who is responsible when junior staff or outside advisers use AI to help prepare documents.
A sensible process uses AI for early research, organisation, and drafting while reserving decisions and verification for accountable people. It preserves sources, versions, and records of corrections. Those controls are especially important where an error could affect someone's rights, money, employment, or reputation.
Sources
- US Court of Appeals for the Fifth Circuit, 22 October 2026 — judgment in case 25-60496.
- Reuters, 22 October 2026 — appeals court and AI errors.
- ISO/IEC 42001 explained — International Organization for Standardization.
We design approval and source-verification procedures for high-impact AI-assisted work — competence, oversight, and audit trails through our gap assessment.