That ChatGPT transcript I rushed through ended up on page 3 of a court filing

I uploaded a conversation from the GPT store into a legal research tool last Tuesday without checking the output first. Turned out the model had recommended a strategy based on a case that was actually overturned in 2019. The judge's clerk caught it during discovery and I looked like an idiot in front of my whole team. Has anyone else accidentally trusted an AI summary without verifying the source law first?

3 comments

3 Comments

phoenix_bailey2d ago

Did your legal research tool have any disclaimer about AI accuracy before you hit upload? I found out the hard way that some of these GPT store models are fine-tuned on reddit threads and blog posts, not actual caselaw databases. The one I tried recommended a motion strategy based on a district court opinion that got slapped down by the circuit court last year. Now I run every single cite through Westlaw before I even mention it in a draft. Takes an extra ten minutes but saves you from that horrible feeling when opposing counsel emails you a string of "actually, that case was reversed" notes.

river9522d ago

Totally feel that. I had a GPT store tool tell me a case was "binding precedent" from the Ninth Circuit. Turns out it was a footnote from a district court in Alaska that got vacated three years ago. Lucky for me I checked it before filing, but that was a close call that made me look like an idiot in front of my team. Now I treat every single AI output like it crawled out of a Reddit argument thread, because honestly that's probably closer to the truth than they'll admit. That extra Westlaw check is the only thing keeping my malpractice insurance from going through the roof.

morgan_king362d ago

Oh man, that's brutal. Honest to god this reminds me of the time I copy-pasted a whole paragraph from a legal blog into a memo and it turned out to be from a parody site. My paralegal still brings it up at happy hour. Tbh I'm not even allowed near the AI tools anymore after that disaster.