1,227 Fake Citations and Counting. Courts Are Done Being Patient.
AI hallucinations in legal filings have gone from cautionary tale to full-blown crisis. The Ninth Circuit just sanctioned two attorneys on June 3, 2026. An Oregon court just set a new record: $110,000. And a DOJ attorney got fired. Here is where this stands.
In May 2023, a judge in the Southern District of New York sanctioned two attorneys for submitting six fabricated case citations generated by ChatGPT. It felt like a landmark moment — a cautionary tale that would reshape how the legal profession approached AI tools. The attorneys were fined $5,000 and the case became required reading in legal ethics courses nationwide.
Three years later, the number of documented AI hallucination incidents in court filings worldwide has reached 1,227 — and it is growing at five to six new cases every day. The $5,000 fine that shocked the profession in 2023 has been replaced by a $110,000 sanction in Oregon and case dismissal in the Sixth Circuit. On June 3, 2026, the Ninth Circuit issued its own sanctions order against two attorneys who submitted briefs containing multiple nonexistent cases, misattributed quotations, and gross misrepresentations of real cases.
Courts are no longer treating this as a novelty. They are treating it as professional misconduct — and the consequences are escalating accordingly.
The Sanctions Are Getting Bigger. The Cases Keep Coming.
The scale of the problem in 2026 is difficult to overstate. Damien Charlotin, a research fellow at HEC Paris's Smart Law Hub who maintains the most comprehensive tracking database of AI hallucination incidents in court filings, cataloged 1,227 cases globally as of early 2026. The number was 200 a year earlier. It was 719 in January 2026.
The most recent judicial responses reflect a profession losing patience. On June 3, 2026, the Ninth Circuit issued sanctions against two attorneys in LNU v. Blanche for filing briefs containing multiple nonexistent cases, misattributed quotations, and gross misrepresentations of real cases. The court was explicit: it was not sanctioning the use of AI. It was sanctioning the failure to verify AI output before signing and filing it. "The peril of abdicating professional care remains with the attorney."
In April 2026, a federal magistrate judge in Oregon imposed $110,000 in sanctions — the largest AI hallucination sanction in United States legal history — and dismissed a $12 million case with prejudice after finding 23 fabricated citations across three court filings. A DOJ attorney was fired for a similar failure. The Seventh Circuit addressed the issue in March 2026 in Dec v. Mullin, declining further sanctions only because the errors were unintentional and counsel was contrite — but admonishing that the responsibility to verify remains entirely with the attorney.
Courts are not punishing lawyers for using AI. They are punishing lawyers for trusting it — and filing what it produces without verification.
This Is Not Just a Problem for the Filing Attorney
The AI hallucination crisis in legal filings is widely discussed as an attorney responsibility issue — and it is. But it carries direct implications for litigation support teams, legal operations, and the organizations those teams serve.
When opposing counsel submits a brief with fabricated citations, your team needs to be equipped to identify it. The attorneys in the $31,100 California case were caught only because a judge noticed the citations were unverifiable. In other cases, fabricated authority has nearly made it into court orders. Legal teams that lack a process for verifying opposing counsel's citations — or for flagging AI-generated content in briefs received — are operating with a blind spot.
For organizations under litigation, the risk is symmetric. If your outside counsel or litigation support team is using AI to assist with brief writing, research, or document analysis, the accountability structure needs to be clear: every citation must be independently verified before it is filed. The 59% of documented hallucination incidents involving pro se litigants reveals the paradox — the people most likely to use AI for legal work without verification are those least equipped to recognize its failures.
What Every Legal Team Needs to Know
Five Things That Define the Crisis in 2026
- 1,227 documented cases and growing at 5–6 per day. This is no longer an edge case — it is a systemic failure in how AI output is being used in legal work.
- Sanctions have escalated from $5,000 to $110,000 in three years. Courts have moved from cautionary warnings to case dismissal and record-setting fines.
- Using AI is not the violation. Filing unverified AI output is. The Ninth Circuit, Seventh Circuit, and Sixth Circuit have all been explicit on this point.
- AI will confirm its own hallucinations. Asking ChatGPT to verify a case it fabricated will result in a confident confirmation. Verification must happen outside the AI tool.
- 300+ federal judges have AI-specific standing orders. If your team doesn't know what the judges in your matters require, that gap is a risk.
The AI hallucination crisis is, at its core, a workflow problem. The technology produces confident-sounding output. Attorneys under deadline pressure use it. The verification step — which requires independent checking outside the AI tool — gets skipped. The result is a fabricated citation in a court filing, and consequences that can end a case and a career.
For legal operations teams and litigation support professionals, the practical response is not to prohibit AI — it is to build the verification step into the workflow with the same discipline as cite-checking has always required. That means knowing which AI tools are being used in brief preparation, building independent verification into the review process, and understanding the standing orders of the judges in each matter. The technology is not going to stop producing hallucinations. The process has to account for them.
The legal profession has always understood that signing a brief is an act of professional accountability. The AI hallucination crisis did not change that rule. It exposed how many attorneys were filing documents they had not fully read.
Courts are not going to lower the verification standard to match the speed of AI. The standard is: if you sign it, you own it. The rest is a workflow problem — and workflow problems can be solved.