A Delaware court ruling handed down in mid-November has drawn sharp attention across the legal profession for demonstrating precisely what separates a protected firm from a punished one.
On November 13, 2026, the Delaware Court of Chancery sanctioned a lawyer who had signed a legal brief containing hallucinated citations generated by artificial intelligence.
The court found the lawyer liable under Delaware’s version of Rule 11, a procedural rule governing the accuracy and good faith of court filings.
Critically, the presiding judge ensured the monetary sanctions came directly from the lawyer’s own pocket, not from the firm’s coffers.
Yet the lawyer’s firm walked away without sanction, shielded by something the court explicitly recognised: a serious, well-constructed AI use policy that was already in place before the incident occurred.
The court pointed to the Delaware Lawyers’ Rules of Professional Conduct, which oblige law firm leadership to take reasonable steps to ensure that ethical rules are followed throughout the organisation.
Judges also referenced ABA Formal Opinion 512, which calls on firms to establish unambiguous policies governing how AI tools may be used by their lawyers and staff.
The firm in question had gone well beyond drafting a policy document, building an internal training programme that tackled AI hallucinations head-on and addressed professional duties “concerning competence, confidentiality, meritorious claims, candor, and supervision.”
That programme earned accreditation from Delaware’s Commission on Continuing Legal Education for one hour of enhanced ethics credit, a recognition that underscored its substantive quality.
The court’s verdict on the firm’s conduct was pointed and unambiguous: “In short, [the firm] did the right thing.”
Legal commentator Stephen Embry, who writes the TechLaw Crossroads blog and works as a speaker and adviser to the legal industry, highlighted the ruling as a defining illustration of what effective AI governance actually demands.
Embry argues that a functional AI policy cannot be treated as a single document to file and forget, describing the requirements as a three-legged stool that firms must build and maintain over time.
The first leg is the policy itself, which must be thorough enough to cover approved tools, confidentiality obligations, and the verification of any AI-generated citations or information.
The second leg is implementation, meaning firms must translate written guidelines into genuine practice through training, supervision, and accountability structures that reach every level of the organisation.
The third leg is continuous updating, because a policy written in 2023, for instance, would almost certainly have failed to anticipate the emergence of AI agents and the distinct risks they introduce.
Embry, who raises these points regularly in presentations to legal professionals, stresses that the pace of AI development means yesterday’s guidelines can become dangerously inadequate without any deliberate decision to abandon them.
The Delaware outcome makes the stakes concrete: firms that invest in living, substantive AI governance frameworks can stand behind their policies when things go wrong in court.
Those that treat AI guidelines as a compliance checkbox, or that have never drawn them up at all, leave both their lawyers and their clients exposed to consequences that are now, as this ruling confirms, very real.

