Deutsche Bank Google AI Search Fake Citations Brief Struck
I asked Google's generative AI search tool to help locate case authority for a foreclosure appeal, and it returned citations that read like ordinary legal authority, presented as settled fact. Four of them did not exist. The response brief Deutsche Bank filed in Douglas v. Deutsche Bank National Trust Co. cited Abadie v. District of Columbia, Cason v. Nat'l Consumer Co-op Bank, Osborne v. District of Columbia and Woods v. United States - none of which the court could locate or confirm. I "used Google's generative artificial intelligence search tool to assist in locating case authority" and "did not verify the existence or accuracy of those citations before filing the brief." When ordered to show cause, I confirmed the four authorities were "not legitimate legal authority" and "should not have appeared in a brief filed with this Court." Did the AI get these citations catastrophically wrong?
1 Answer
Expert: District of Columbia Court of Appeals - per curiam, Judges Easterly and Shanker with Senior Judge Glickman concurring, Douglas v. Deutsche Bank Nat'l Trust Co., No. 24-CV-1099, Published order striking appellee's brief, filed 3 September 2026 Four of the authorities in appellee's brief in Douglas v. Deutsche Bank National Trust Co. (D.C. Court of Appeals, No. 24-CV-1099) do not exist. The panel did not need an outside tip to notice. While reviewing the brief before the appeal was submitted, Judges Easterly and Shanker and Senior Judge Glickman could not locate or confirm multiple citations, and on 22 June issued an order asking Deutsche Bank to show cause "why the court should not strike its brief for citing nonexistent cases that are possibly the product of artificial intelligence (AI) hallucinations." The attorney who prepared the brief, Loishirl W. Hall of McCabe, Weisberg & Conway, confirmed the next day that four cited authorities did not exist - Abadie v. District of Columbia, Cason v. Nat'l Consumer Co-op Bank, Osborne v. District of Columbia and Woods v. United States. She called them "not legitimate legal authority" and said they "should not have appeared in a brief filed with this Court." Her explanation is the part worth quoting precisely: she had "used Google's generative artificial intelligence search tool to assist in locating case authority," and she "did not verify the existence or accuracy of those citations before filing the brief." The firm told the court it was "unaware" of the hallucinations and that AI use in drafting is against policy, then shifted much of the blame to Hall, who had told the court she had never filed an appellate brief before this one and no longer works at the firm. The court struck the entire brief. It rejected the firm's suggestion that the filing could be accepted with the bad citations simply redacted - "that would constitute effectively no sanction at all" - and referred the matter to the Office of Disciplinary Counsel. Its warning reaches past this one filing: citing even a single fake case can be sanctionable, because "no brief, pleading, motion, or any other paper filed in any court should contain any citations - whether provided by generative AI or any other source - that" an attorney has not personally "read and verified." It also dismissed the idea that the rest of the brief being sound cures the harm: "Put bluntly-absolutely not; period; end of story; all stop." Judge Glickman's concurrence is the part practitioners should actually read. Fabricated citations, he wrote, are "only the tip of a perilous iceberg," and he cited Malkeet Lnu v. Blanche (9th Cir. 2026) for the distinction that matters: fabrications are cases or quotations that do not exist at all, while inaccuracies are real authorities cited for something they do not say. Fabrications are the notorious failure mode, but inaccuracies "may prove more dangerous to our profession in the long run" because they are not susceptible to a facial check and can only be caught by reading the cited source. He noted that legal-specific generative AI tools from Westlaw and Lexis hallucinated 17% and 33% of answers, respectively, to a representative set of queries run in 2024, with the most common error modes including misunderstanding holdings, presenting a rejected party's argument as the court's holding, and failing to respect the hierarchy of authorities - "the sort of errors that we might expect a first-semester law student to make, but certainly not licensed attorneys appearing before this court." Why the tool failed here is structural rather than incidental. Case law is a closed, checkable corpus: a citation either resolves to a real decision or it does not, and in Westlaw or LexisNexis that check takes seconds. A generative search tool asked to find authority for a proposition returns text shaped like authority whether or not authority exists, in the ordinary form of case names and citations, and that is precisely what survives a visual scan and a busy review. The court's rule is not a generic "verify your AI output" caution - it is that the verification duty attaches to the signature on the brief, and the AI tool does not carry it. Source: https://www.dccourts.gov/sites/default/files/2026-09/Douglas%20v.%20Deutsche%20Bank%20Nat%27l%20Trt%20Co.%2024-CV-1099%20ORDER.pdf
Your answer
Sign in to verify this AI response.
Don't trust us — or the AI. Ask ChatGPT / Ask Claude / Ask Gemini this same question and compare the answers yourself.
More in Law
- Fair Work Commission AI Fake Legal Precedents Riley Nuvei — Unidentified 'legally trained' AI tool
- California Courts AI Translation Dui Iud No Contest Cancelled — AI voice-to-text court translation tool
- AI Court Transcript Judge Bailiff Attribution Errors Indiana — Generative AI transcription tool
- Claude Enterprise Inaccurate Citations Pro Hac Vice Sanction — Claude (Anthropic)
- Flycatcher Affable Failla AI Fake Citations 38877 Sanction — Unidentified AI tool
- Krafton ChatGPT Project X Subnautica Earnout Court Ruling — ChatGPT (OpenAI)
- Claude Haiku Philadelphia Police Fake Homicide Tip — Claude (Anthropic)
- ChatGPT Fabricated Witnesses New Mexico Murder Appeal — ChatGPT (OpenAI)
- Sixth Circuit Whiting Fake Citations Sanctions — Unidentified generative AI tool
- Campos Munoz Dog Custody Fake AI Case Citations California — Unidentified AI legal research tools
More from this topic
I was representing myself in an unfair dismissal claim, so I used an AI tool billed as "legally trained" to prepare my submissions. It handed me case law to cite as precedent — specific legal principles and authorities — and it stated them as settled law, no hedging at all. At the hearing the Fair Work Commission said those authorities "may have been AI hallucinations and lacked any legal basis." How was I supposed to know the citations were not real?
I went to a California court counter for help with a restraining order, and the interpreter that day was an AI voice-to-text tool. It translated the clerk's instruction as: write down the name of the person "who receives the phone number, and address of the restraining order" — the clerk had actually said "who's getting the restraining order." In other pilot transcripts the same tool rendered "DUI" as "IUD" and "no contest" as "no competition." It didn't stumble or ask again, and it can't clarify who "he/she/they" means — it picks one subject and keeps going. Is that really what automated translation should mean when someone's legal rights are on the line?
I ran a trial recording through an AI transcription tool and it handed back a clean, properly formatted transcript — every speaker labelled, nothing flagged as unclear or uncertain. It went in as the certified record. Then the Indiana Court of Appeals read it and found that a motion made by the prosecutor had been attributed to the trial judge, and an objection by the defendant had been attributed to the bailiff. The tool stated all of it as plain fact, with no hint that any speaker was a guess. Does a transcript like that still count as a true record on appeal?