The Justice Department's ethics office found one lawyer misled a court about AI use and another uploaded protected files to a public AI tool.
Most AI misconduct stories so far have involved private lawyers and fake citations. This week the Justice Department published findings against two of its own attorneys, and only one of them is about citations. The other is about where confidential documents went.
On November 17, the department’s Office of Professional Responsibility posted two anonymized investigative summaries, [Summary 5]( https://www.justice.gov/opr/2026-investigative-summary-5) and [Summary 6]( https://www.justice.gov/opr/2026-investigative-summary-6). In both, OPR concluded the attorney violated the rules of professional conduct and referred its findings to the department’s Professional Misconduct Review Unit.
The Cover Story Was the Bigger Problem
According to Summary 5, an Assistant U.S. Attorney filed a pleading that misrepresented case holdings and included fabricated quotations. After the court ordered the government to explain, he filed a brief saying the errors came from accidentally filing a draft. He did not tell the court the errors came from his use of an AI platform. At a later hearing, OPR says, he made false and misleading statements to the court about his use of generative AI.
OPR found he knowingly and intentionally violated his duty of candor on both counts: the explanatory brief and the hearing. The original errors are not where the finding lands. The finding is about what he told the court afterward.
[Bloomberg Law]( https://news.bloomberglaw.com/us-law-week/use-of-ai-leads-two-doj-lawyers-to-face-potential-bar-discipline) reported that this summary matched its earlier reporting on a North Carolina Assistant U.S. Attorney who was fired in March and later reprimanded by a judge. OPR itself names no one.
A Confidentiality Case, Not a Citation Case
Summary 6 involves an attorney in a department component who used a public AI platform for government work without approval. He was detected uploading a document whose header showed it was covered by a tribunal’s protective order. Later reviews found more uploads: a second document under the same order and documents containing department employees’ personally identifiable information, including medical information.
OPR reached three conclusions. Using the unapproved public platform in violation of department policy was a reckless violation of the professional conduct rules. Uploading confidential department documents was knowing and intentional. Uploading the two protected documents knowingly violated both the conduct rules and the tribunal’s order.
Nothing in Summary 6 says the AI produced a wrong answer. The violation was the upload itself. That is the risk most firms and legal departments can actually control, through approved tool lists, data handling rules and monitoring.
What Happens Next
OPR investigates but does not impose discipline. Under the department’s [published process]( https://www.federalregister.gov/documents/2026/03/05/2026-04390/review-of-state-bar-complaints-and-allegations-against-department-of-justice-attorneys), the Professional Misconduct Review Unit reviews OPR’s findings, decides whether they are supported by a preponderance of the evidence, and decides on discipline, which can include reprimand, suspension or termination. When a state rule of professional conduct is implicated, it can authorize OPR to refer the matter to bar disciplinary authorities.
The same week, the European Court of Human Rights [announced]( https://www.echr.coe.int/w/romanian-lawyer-barred-for-two-years-from-representing-applicants-before-the-court-over-unverified-ai-generated-citations) on November 18 that its President had barred a Romanian lawyer from representing applicants before the court for two years. A Chamber found her observations cited nonexistent cases and fictitious paragraphs from real judgments. She acknowledged using AI tools to draft them.
The Chamber found “a pattern of numerous, repeated, substantive errors
rather than isolated citation mistakes,” the court’s press release said.
Common Questions
Were the DOJ lawyers named?
No. OPR’s summaries are anonymized and do not identify the courts, cases or AI tools involved.
Have they been disciplined?
Not yet, based on the public record. OPR referred its findings to the Professional Misconduct Review Unit, which decides discipline and whether to authorize a bar referral.
Was the AI tool itself the problem in the second case?
No. OPR’s findings rest on using an unapproved public platform and uploading protected and confidential documents to it.
Does this apply outside government?
The duties at issue, candor to the tribunal, confidentiality and obeying court orders, apply to every lawyer. The DOJ findings show how an investigator breaks them down.
What to Watch
Whether the review unit authorizes bar referrals, and whether more agencies publish AI use findings. For legal ops teams, Summary 6 is the one to circulate: an approved tool list only works if someone can see when people go around it.
[image: Legal Tech Circle] <https://legaltechcircle.com/> Quinn Ashby The Circle Desk, Legal Tech Circle quinn@legaltechcircle.com | legaltechcircle.com Legal Tech Funding Tracker <https://legaltechcircle.com/market-signals/funding-tracker/?utm_source=email&utm_medium=signature&utm_campaign=quinn>: every round since 2012
In this story
See every round in the Funding Tracker