A federal judge in Oregon imposed USD 110,000 in fines and attorney fees against two lawyers who filed court documents containing fabricated case citations and invented quotations generated by AI. US Magistrate Judge Mark D. Clarke of the District of Oregon wrote that “in the quickly expanding universe of cases involving sanctions for the misuse of artificial intelligence, this case is a notorious outlier in both degree and volume.”
The case involved a winery inheritance dispute. Across three separate briefs filed over five months, the plaintiffs’ legal filings contained 15 fabricated case citations and eight invented quotations produced by AI. When the fake cases were identified, the lawyers did not explain how the citations were generated. They offered to file a notice of errata if the defendants pursued a motion for sanctions. The court found that the lawyers took no steps to verify the citations using a conventional citator and noted the “total lack of remorse” from lead counsel. The court dismissed the plaintiff’s claims with prejudice, imposed the sanctions, and referred the matter to the Oregon State Bar.
The penalty is believed to be the largest AI-related sanction in US legal history. It is not an outlier in a trend that is accelerating.
The Q1 2026 sanction wave
ComplexDiscovery reported that US courts imposed at least USD 145,000 in sanctions for AI-generated fabricated citations in the first quarter of 2026 alone. Sanctions reportedly stood at USD 5,000 in January and USD 250 in February before March brought a sudden surge, including a USD 30,000 penalty from the Sixth Circuit and the Oregon cases.
Damien Charlotin of HEC Paris’s Smart Law Hub, who maintains a global database of AI hallucinations in legal proceedings, told NPR that the problem is appearing with “unsettling frequency,” with multiple courts flagging fabricated filings on the same day.
OPB reported that according to Ankur Doshi, general counsel of the Oregon State Bar, there have been approximately five identified court filings in Oregon containing AI hallucinations. Nationally, Doshi said, that number is closer to 900.
The Oregon Court of Appeals issued a rare public warning from Chief Judge Erin C. Lagesen, stating that AI-fabricated filings are “rapidly escalating” and coming from “lawyers and self-represented litigants alike.” Lagesen directed court staff and judges to track the time spent addressing fabricated authority “so that Oregonians can have an accounting of the resources consumed.”
The Oregon sanction formula
Oregon has moved from case-by-case sanction decisions to a structured formula. The Oregon Court of Appeals established a per-item tariff in late 2025: USD 500 per fabricated citation and USD 1,000 per fabricated quotation. Federal courts in Oregon have followed the pattern.
The formula changes the risk calculus for every lawyer and legal team using AI for research or drafting. AI hallucination sanctions are no longer unpredictable judicial discretion. They are quantifiable costs that scale linearly with the number of fabricated items. A brief containing 10 fake citations and 5 fake quotations carries a baseline sanction exposure of USD 10,000 before attorney fees, case dismissal, or bar referral are added.
The Oregon Court of Appeals has already applied the formula multiple times. A USD 10,000 fine in March involved a marijuana production licence case where the court found at least 15 fabricated case citations. An approximately USD 8,000 fine followed a brief containing fabricated quotations and propositions of law falsely attributed to existing cases.
A Northwestern University survey suggests more than 60% of federal judges are now using AI tools themselves, creating a sharp divide between the standards courts demand of lawyers and the technology practices inside the judiciary. The gap is likely to close as courts establish clearer rules, but for now, judges are sanctioning lawyers for the same tools some judges use.
Why this matters beyond law
The legal profession is the first sector where AI governance failures carry immediate, documented, escalating financial penalties. Other professions are approaching the same threshold.
SAW covered the Australian Fair Work Commission’s response to AI-generated claims in April. The FWC found its workload up 70% partly due to AI-generated applications, some containing fabricated elements. The FWC’s draft guidance note addresses the issue from the Australian side. The Oregon sanctions address it from the US enforcement side. The pattern is the same: AI tools are producing plausible-looking professional work product that contains fabricated content, and the people submitting it are either not checking or not equipped to check.
The Novorésumé finding that 53% of workers are not transparent about their AI use and 1 in 6 have been promoted on AI-produced work connects directly. If employees are submitting AI-generated work without disclosure across industries, the legal profession is simply the one where fabricated content gets caught fastest because citations are verifiable. In accounting, consulting, engineering, and healthcare, AI-generated work product may contain errors that take longer to surface but carry equal or greater consequences.
Sullivan and Cromwell, one of the world’s largest law firms, filed an emergency letter on 18 April 2026 asking a bankruptcy judge in the Southern District of New York to avoid sanctions after admitting that its court filing contained AI-generated hallucinations, including fabricated citations and legal errors. If a firm of that calibre can submit fabricated AI content, the risk applies everywhere.
What legal teams and in-house counsel should do
Implement a mandatory verification workflow. Every AI-assisted legal document must be checked against a conventional citator (Westlaw, LexisNexis) before submission. The Oregon federal court explicitly noted that the sanctioned lawyers “took no steps to verify the citations.” The verification step is the control. Without it, the organisation is accepting the risk that any AI-generated citation could be fabricated.
Define approved AI tools for legal research. The Oregon cases involved clients using unidentified AI tools and passing the output to lawyers. Legal teams need an approved tool list that specifies which AI services may be used for research and drafting, with explicit prohibitions on submitting AI-generated content that has not been through the verification workflow.
Log AI use in legal work product. The Oregon sanctions escalated partly because the lawyers could not explain how the fabricated citations were generated. Logging which AI tools were used, when, and by whom creates an audit trail that protects the team in case of errors and demonstrates good-faith compliance with emerging court expectations.
Update outside counsel guidelines. In-house legal teams should update their panel firm guidelines to require disclosure of AI tool use in work product, verification workflows for AI-assisted research, and explicit responsibility for the accuracy of all citations regardless of how they were generated. The Littler survey SAW covered on 15 May found 79% of employers expect AI-related litigation in the next 12 months. Outside counsel governance is where that expectation meets operational reality.
Brief the board on professional liability exposure. AI-generated errors in legal, accounting, and consulting work product are a professional liability risk, not just a technology risk. If the organisation’s professional indemnity insurance covers AI-assisted work, the insurer may require evidence of verification workflows and approved tool policies. If the insurance does not explicitly cover AI-assisted errors, that gap needs to be identified and addressed before a claim arises.
Sources
- ABA Journal, “Federal judge hands down $110K penalty against 2 lawyers for AI errors in court documents,” 16 April 2026 (Judge Clarke opinion, $110K penalty, “notorious outlier” quote, winery case details). abajournal.com
- OPB, “AI fabrications in legal filings grow in Oregon, US,” 10 May 2026 (Doshi quote, five Oregon filings, 900 nationally, OPB investigation). opb.org
- Oregon Capital Chronicle, “AI-generated erroneous court filings are ‘rapidly escalating,’ Oregon appeals judge warns,” 30 April 2026 (Chief Judge Lagesen warning, $10K and $8K fines, time-tracking directive). oregoncapitalchronicle.com
- NWSidebar (Washington State Bar), “Parade of Horribles: Federal Court in Oregon Surveys Sanctions for AI Fake Citations,” 2 March 2026 (15 fake citations, 8 fabricated quotations, no citator verification, “total lack of remorse”). nwsidebar.wsba.org
- ComplexDiscovery, “The AI Sanction Wave: $145K in Q1 Penalties,” 5 April 2026 (Q1 2026 tally, January-March trajectory, Charlotin database, Sixth Circuit penalty). complexdiscovery.com
- Helsell Fetterman, “AI Hallucinations Keep Costing Lawyers in Court,” 23 April 2026 (winery case detail, Sullivan & Cromwell emergency letter, bar referral). helsell.com
- KGW Portland, “Oregon winery dispute leads to major sanctions over AI-generated court filings,” April 2026 (largest sanction in US, KGW investigation, Gustitus quote). kgw.com