Part of the AI Regulation News hub.
Delaware's Court of Chancery made a law firm, not just its partner, answer for AI-fabricated citations
A brief carried fictitious citations and fabricated quotations. What makes the order worth reading is not the sanction, because there was none. It is that the firm had to hand over its AI policies.
Bottom line: A Rule to Show Cause, not a sanction. On the record we opened, no penalty had been imposed and the question was left open.
Who this affects: Corporate and M&A litigators, Delaware counsel, and the law-firm general counsel and risk partners who own GenAI policy.
Effective date: Letter ruling dated 1 July 2026. Affidavits were due no later than 15 July 2026.
What changed: The court invoked Rule 11(c)(1) to make the firm answer alongside the signing partner, and ordered disclosure of the firm's written GenAI policies and training programmes.
Analysis: Sanctions rulings against individual lawyers have piled up over the past two years. What separates this order is that the court required the firm itself to address its written GenAI policies, guidelines, restrictions and training in answer to a Rule to Show Cause.
Primary sources: Delaware Court of Chancery letter ruling (PDF)
- Instrument (EN)
- Letter ruling and Rule to Show Cause, Leiske v. Kidd, C.A. No. 2025-0426-CDW (LWW)
- Authority
- Court of Chancery of the State of Delaware, Vice Chancellor Lori W. Will
- Jurisdiction
- United States, Delaware
- Status
- Rule to Show Cause issued. No sanction imposed on the record opened.
- Bindingness
- Binding order on the parties. Applies existing Court of Chancery Rule 11, not a new AI rule.
- Effective / next deadline
- Decided 1 July 2026. Affidavits due no later than 15 July 2026.
- Primary source
- https://courts.delaware.gov/opinions/download.aspx?id=397720
What happened
The underlying case is an advancement dispute that was before a Magistrate in Chancery. It was reassigned to Vice Chancellor Lori W. Will "for the limited purpose of resolving exceptions to the Magistrate's final report on advancement."
The plaintiffs' answering brief, in the court's words, "contained fictitious citations, fabricated quotations, and hallucinated legal propositions."
On 28 January, plaintiffs' counsel filed a letter "acknowledging the use of a generative artificial intelligence ('GenAI') tool to revise the text of the Answering Brief," and "admitted failing to verify the accuracy of the citations before filing, attributing the oversight to a paralegal's review."
A corrected brief followed. The court noted it "removed quotation marks around erroneous statements of law" but "did not correct the inaccurate descriptions." The next day counsel faulted opposing counsel for not conferring first, maintained that the cited authorities "are real" and "support the arguments advanced," and accused the defendants of using the errors to gain a "litigation advantage."
On 2 February the court overruled the exceptions and expressly kept the remedy question alive: "I retain jurisdiction to determine what remedy, if any, should issue to address the GenAI errors in the plaintiffs' brief."
The holdings
Court of Chancery Rule 11(b) requires an attorney presenting a filing to certify that the legal contentions made "are warranted by existing law." The court treated that signature as substantive: an attorney's signature "is not a procedural formality; it is a certification to the tribunal."
On the tool itself, the court was measured. Using GenAI in court filings is "not inherently problematic," provided the attorney or litigant "carefully verifies the accuracy of the output." That language is quoted from an earlier Chancery decision, An v. Archblock, Inc.
Then the operative rule: "The submission of hallucinated legal propositions raises an inference of a Rule 11(b) violation, requiring counsel to show cause why sanctions should not issue." And the reason why, stated in six words: "A fictitious citation is not 'existing law.'"
On delegation, the court's language was direct. "The duty of candor to the court, and the obligation to conduct a reasonable inquiry under Rule 11, are nondelegable. Counsel cannot discharge those obligations by relying on a paralegal, much less on an artificial intelligence program."
Why the firm-level holding is the real story
Rule 11(c)(1) provides that "[a]bsent exceptional circumstances, a law firm must be held jointly responsible for violations committed by its partners, associates, or employees." The court applied it here on an explicit rationale: "because this incident may implicate the firm's training, supervision, and deployment of GenAI, the firm must answer alongside the individual signatory."
That is a different proposition from the sanctions cases that have accumulated over the past two years. Those ask whether a lawyer verified a citation. This asks whether a firm built a system in which the citation would be verified.
The order reflects it. The individual's affidavit had to address how GenAI was used in drafting, "including who entered the prompts and how the generated output was incorporated into the Answering Brief," what instructions were given to the paralegals, whether attorneys verified the cited text, and what remedial steps were taken.
The firm's affidavit had to address "[t]he firm's written policies, guidelines, and restrictions regarding the use of GenAI by attorneys and staff (including paralegals)," the training programmes it has implemented or plans to implement, and why it should not be held jointly responsible.
That is not discovery. It is a Rule 11(c)(1) affidavit requirement, and it obliged a firm to set out its AI governance programme in a court filing within two weeks.
What is still open
No sanction had been imposed. The court said it would decide, on receipt of the affidavits, whether an evidentiary hearing or oral argument was necessary "or whether the Rule to Show Cause will be resolved on the papers."
Affidavits were due no later than 15 July 2026. We have opened no record of what was filed or how the court resolved it, and we are not going to guess. The matter is unresolved on the record available to us, and anyone relying on this piece should treat the outcome as pending.
The court also did not identify which GenAI product was used. It is not in the order, so it is not here.
The practical consequence
Three things follow for a firm of any size.
A firm's written AI policy can become part of a court-ordered Rule 11 response. If yours is thin, aspirational or undated, that is the version you may have to explain to a court. The time to fix it is before a rule to show cause, not after.
The court's stated rationale for reaching the firm was its training, supervision and deployment of GenAI. A policy that nobody was trained on would read badly against that sentence.
The court treated the response as compounding the original error: deleting quotation marks while leaving the fabricated legal synthesis in place, then arguing about meet-and-confer etiquette. The court's answer to that was direct. "There is nothing to negotiate or compromise when an attorney submits false citations to a tribunal."
Nobody was sanctioned here. What the court did was require a firm to set out its written GenAI policies and training in a Rule 11 affidavit within two weeks, so treat your AI governance programme as something that may have to be filed.
Source File
https://courts.delaware.gov/opinions/download.aspx?id=397720
Open the Delaware Courts PDF and confirm the 1 July 2026 date, the caption Leiske v. Kidd, C.A. No. 2025-0426-CDW (LWW), and the Rule to Show Cause section setting the 15 July 2026 affidavit deadline.
The submission of hallucinated legal propositions raises an inference of a Rule 11(b) violation, requiring counsel to show cause why sanctions should not issue. A fictitious citation is not 'existing law.' ยท Vice Chancellor Lori W. Will, 1 July 2026
FAQ
Was anyone sanctioned?
No. The court issued a Rule to Show Cause requiring counsel and the firm to explain why sanctions should not be imposed, and set an affidavit deadline of 15 July 2026. On the record we opened, no sanction had been imposed and the question was left open.
What exactly did the court order?
Affidavits from the signing partner and from the firm. The partner had to explain how GenAI was used, including who entered the prompts and how the output was incorporated. The firm had to disclose its written policies, guidelines and restrictions on GenAI use by attorneys and staff, and its training programmes.
Why does it matter that the firm had to answer?
Because Rule 11(c)(1) holds a firm jointly responsible for violations by its partners, associates or employees, and the court reasoned that the incident may implicate the firm's training, supervision and deployment of GenAI. That moves the inquiry from one lawyer's verification failure to the firm's governance.
What should a firm do about its AI policy now?
Assume it may have to be disclosed to a court in a Rule 11 response. Make sure it is written, dated, specific about verification duties, and actually delivered as training rather than circulated as a memo. The court's stated rationale was training, supervision and deployment.
Related briefings
Sponsored Training
Practical AI training for regulated professionals, built around verification, documentation and a defensible process. See the courses.