Building an International AI-Hallucination Case Tracker and Why I’m Sharing It Half-Finished!
I’ve spent a long time collecting court or tribunal decisions where lawyers, litigants, experts or AI tools themselves have put non-existent authorities or principles before the court. The phenomenon has jumped borders as quickly as the technology: from Mata v Avianca in New York to Nexgen Pathology v Darceuil Duncan in Trinidad & Tobago and Reclamação 78.890 in Brazil.
The aim is to create a live, international tracker of these decisions, free to anyone who needs to understand how courts are reacting to “AI hallucinations”.
Why the tracker matters
It’s a practical compliance tool. Anyone can see, at a glance, the sanctions other courts are imposing.
Comparative perspective. Trends are clearer when you line up the UK’s costs orders beside U.S. referral threats and Caribbean disciplinary referrals.
Policy evidence base. It assists people considering policies on this issue by linking important data.
Why it’s going out now, warts and all
I have had some really interesting discussions about many of these cases, and people have asked to see the extent of the problem. Now that I’m in court, I find it more difficult to publish my analysis as quickly as I’d like. My client work keeps pushing publication to the bottom of my list, so I’ve stopped waiting for a ‘perfect’ launch.
I’ve thrown the raw research and my notes into a publicly viewable table on Natural & Artificial Law:
https://naturalandartificiallaw.com/ai-hallucination-cases-tracker/
The list almost certainly omits decisions and may contain transcription errors. Nonetheless, the community gains more by seeing the data today than by waiting for me to finish a polished report.
How you can improve it
If anyone is interested in assisting me in this project, I would greatly appreciate help with the following:
Peer-review the entries. Please flag typos, wrong dates, duplicate rows, anything.
Send missing cases. Domestic rulings, professional-disciplinary findings, unpublished orders… every jurisdiction helps.
Suggest new fields. You may find further information helpful in the table. Let me know!
Add comments beneath the tracker, reply on Substack, or email me directly. Collaboration will turn the table into a first-rate, crowd-validated resource.
AI tools (yes, including ChatGPT) accelerated the data-gathering stage, but every citation still requires manual verification. Collective vigilance will keep the tracker accurate.
Let’s turn scattered stories into a dependable, global reference!