Background
This advancement case was pending before a Magistrate in Chancery when Vice Chancellor Lori W. Will took jurisdiction to resolve exceptions to the Magistrate’s final report. On January 22, 2026, plaintiffs’ counsel filed an Answering Brief opposing defendants’ exceptions. Two days later, defendants identified that the brief contained fictitious citations, fabricated quotations, and hallucinated legal propositions—errors created through the use of generative AI.
On January 28, plaintiffs’ counsel acknowledged using a GenAI tool to revise the brief and admitted failing to verify citations before filing, attributing the oversight to inadequate paralegal review. Although counsel filed a corrected Answering Brief, it merely removed quotation marks around the erroneous statements while preserving the fabricated legal synthesis. Defendants subsequently flagged remaining misrepresentations. The court overruled defendants’ exceptions on February 2 but retained jurisdiction to address potential sanctions for the GenAI misuse. On April 2, the Magistrate resolved an unrelated contempt and sanctions motion, leaving the issue of plaintiffs’ counsel’s GenAI errors ripe for decision.
The Court’s Holding
The court held that while using GenAI to assist in legal work is “not inherently problematic,” attorneys must carefully verify the accuracy of all output before filing. The duty to verify citations and conduct a reasonable inquiry under Court of Chancery Rule 11(b) is nondelegable—counsel cannot discharge this obligation by relying on a paralegal or an artificial intelligence program. Fictitious citations do not constitute “existing law” and their submission raises an inference of Rule 11(b) violation, shifting the burden to counsel to show cause why sanctions should not issue.
The court emphasized that counsel’s response to identified errors compounded the misconduct. By merely removing quotation marks while preserving fabricated legal propositions, counsel failed to cure the underlying Rule 11 violations. The court also rejected plaintiffs’ argument that defendants should have met and conferred before alerting the court—the meet-and-confer requirement exists to resolve disputes between parties, not to negotiate about false citations submitted to the tribunal. Additionally, under Rule 11(c)(1), the law firm (Richards, Layton & Finger, P.A.) must be held jointly responsible for violations committed by its partners and employees unless exceptional circumstances exist.
Key Takeaways
- GenAI use in legal practice requires meticulous verification of all citations and legal propositions before filing; unverified output creates Rule 11 liability.
- The attorney’s nondelegable duty to verify citations cannot be satisfied through paralegal review or reliance on AI output; personal attorney verification is required.
- Hallucinated citations and fabricated legal propositions constitute violations of Rule 11(b), and correction by removing quotation marks without addressing the underlying false statements does not cure the violation.
- Law firms are jointly responsible for Rule 11 violations committed by their attorneys and staff, including GenAI-related misconduct, and must implement policies, procedures, and training to prevent such errors.
- Opposing counsel may bring false citations directly to the court’s attention without first meeting and conferring with offending counsel.
Why It Matters
This decision establishes important guardrails for the use of generative AI in legal practice. While the court acknowledges GenAI’s utility in streamlining legal work, it makes unambiguously clear that attorney responsibility for accuracy is absolute and cannot be delegated to paralegals or machines. The holding imposes strict verification obligations as a condition of using GenAI, directly anchoring those obligations to Rule 11’s requirement of reasonable inquiry. Attorneys who file briefs containing hallucinated citations face Rule 11 sanctions regardless of their intent or the tool used.
The decision also holds law firms accountable for GenAI misuse across their organizations. Firms must implement written policies governing GenAI deployment, communicate those policies to all users, provide training, and establish technological safeguards. The court’s rejection of the meet-and-confer objection signals that opposing counsel need not negotiate with adversaries about false citations, reducing any pressure to stay silent about accuracy problems. By issuing a Rule to Show Cause directing detailed affidavits about the firm’s GenAI governance and the specific circumstances of the error, the court has signaled that this issue will result in significant sanctions absent exceptional justification.