Cite Unseen: Courts requiring lawyers to certify AI-cited cases actually exist

Within the last year, a serious problem has emerged in the legal system: lawyers submitting briefs and other documents containing hallucinated cases, quotes and made up legal arguments.

As courts becomes increasingly frustrated by the problem, we are seeing amendments to the rules of court to make very clear what the lawyer on record for a matter has to deliver up to the court.

ChatGPT - Double check my citations! One of the most widely reported instance remains the case of Mata v. Avianca, Inc., in which New York lawyers used ChatGPT to draft a motion that cited multiple cases that did not exist and including fabricated quotations from those cases. When opposing counsel could not locate the decisions, the submitting lawyers asked ChatGPT to confirm their existence. It promptly did.Judge P. Kevin Castel imposed a US$5,000 fine and found the lawyers had acted with "subjective bad faith."

Injurious hallucination

In Wadsworth v. Walmart Inc, three attorneys at Morgan & Morgan, one of the largest personal injury firms in the United States, were sanctioned after filing motions containing eight non-existent cases out of nine cited, generated by the firm's own in-house AI platform. Judge Kelly Rankin revoked the lead attorney's pro hac vice admission (a right to be admitting in one state when the lawyer is only admitted to practice in another state) and imposed a combined US$5,000 in fines, holding that signing a legal document is a "nondelegable" duty which cannot be left to AI.

From wine to vinegar

The largest US sanction to date was in the decision of Courvette v Wisnosky where a raft of fake cases and quotations had been cited in a trust dispute over a winery and the plaintiffs was found to be responsible alongside their lawyers. In applying a fine of $500 per false case and $1,000 per false quotation against the various individuals and lawyers involved, with the Judge saying:

In the quickly expanding universe of cases involving sanctions for the misuse of artificial intelligence, this case is a notorious outlier in both degree and volume. Plaintiffs and their counsel have not been adequately forthcoming, candid, or apologetic about their conduct. If there was ever an “appropriate case” to grant terminating sanctions for the misuse of artificial intelligence, this is it

and

With a known price tag, an attorney could decide that the risk paying $500 per non-existent case or $1,000 per fabricated quotation is outweighed by the potential reward of getting away with outsourcing to a chatbot the time intensive work of legal research and writing.

Generative Errors These are not mere isolated incidents. French academic Damien Charlotin maintains a public tracker of AI hallucination cases in court filings which as at June 2026 records more than 1,597 cases globally across jurisdictions:

• The United States (1,114 cases),

• Canada (134 cases),

• Australia (74 cases),

• The United Kingdom (59 cases), and

• Israel (53 cases).

  
  

The volume makes clear this is a systemic issue across the legal system, impacting both lawyer's use of AI as well as self-represented litigants.

"I see dead cases, walking around like regular cases"

What makes AI hallucination in legal filings particularly difficult is that large language models do not distinguish between real and invented authority — they generate plausible-sounding citations with great confidence regardless of whether the underlying case exists.

The Mata lawyers compounded the original error by asking ChatGPT to confirm the cases were real and treating that confirmation as verification, without going to the primary source to check the case.

Court's Pushing Back: Cayman Islands

In October last year, the Cayman Island Court of Appeal gave a clear warning to parties using AI in Samuel Johnson v Cayman Islands Health Service where the court said:-

Those who use artificial intelligence to conduct legal research ... have a professional duty therefore to check the accuracy of such research by reference to authoritative sources, before using it in the course of their professional work (to advise clients or before a court, for example)

Australia - NSW In Australia, the NSW Supreme Court has gone so far as to issue a whole practice note on AI usage, which noted that:

• Material obtained by compulsory means (discovery etc) should not be fed into AI systems outside the control of a law firm;

• In relation to documents to be tendered to the court:

AI must not be used in generating the content of affidavits, witness statements, character references or other material that is intended to reflect the deponent or witness’ evidence and/or opinion, or other material tendered in evidence or used in cross examination

Florida

The Florida Supreme Court has recently moved to amend their rules of court. The rule change does not require disclosure of whether AI was used in drafting, but makes accurate citation an express condition of the filing of submissions with the court, with courts authorised to impose sanctions including reprimand, contempt, striking of the document, dismissal, costs, and attorneys' fees where the rule is breached.

The amendment expands the existing certification in Rule 2.515(d)(2) of the Florida Supreme Court so that a signer of any court filing (whether a lawyer or self-represented litigant) represents that:

"the legal authorities identified exist and are accurately cited."

What's Next?

The continued guidance from courts and rule updates continue a strong trend, so all lawyers should ensure they:

• Verify every cited authority independently before signing any filing or submitting documents to court. AI-generated drafts must be cross-checked carefully against the primary material;
  

• Ensure that in-house or third-party AI tools used for legal research or drafting are understood by all lawyers to require citation verification as a mandatory step, not a recommendation or something to be done if there is time;
  

• Review supervision protocols for junior lawyers and paralegals who may use AI tools without partners reviewing citations before documents are issued, the principals of a law firm are ultimately responsible for what their staff do;

  
  

• Don't wait for a local rule: similar certification and accuracy obligations are likely to be adopted in other jurisdictions and will inform equivalent court guidance; and

  
  

• Document verification steps — in the event of a concern being raised, a clear audit trail will be needed in a matter file to demonstrate that citations were independently confirmed will be material to any defence. Checklists creating records in files of the review will be important.

It is clear that lawyers, who may be used to delegating citation preparation to junior lawyers, are not able to treat AI tools like those junior lawyers, but as has always been the case, will remain responsible, as officers of the court, to ensure the courts are not misled with incorrect information. By Michael Bacina