On the legal privilege issue, small correction -- you are quoting from the prosecution's broadly drafted motion, not the Court. The judge rule from the bench and didn't say much (though we he did say wasn't great -- now same as a well considered written opinion though). If interested, there is some good discussion following my LinkedIn post linked below. Really love your hallucination tracking project too! (Used it as a jumping off point on why lawyers should use LLMs for legal research anyway.... sometimes, and carefully).
Not sure if you saw the comments, but here more for the Gell-Mann amnesia bit. Someone asked Schumer about AI legal hallucinations and his response was “Use the most recent models (try GPT-5.2 Pro in ChatGPT, GPT-5.3-Codex in Codex, Opus 4.5 in Claude) and you won't have this issue.” Hard to imagine someone is that unaware of the limits of LLMs.
Yes, I saw that too ! That's part of the broader credibility point I am making in the footnote, he did not strike me as anyone worth listening to. But he hit a nerve and went viral nonetheless.
On the legal privilege issue, small correction -- you are quoting from the prosecution's broadly drafted motion, not the Court. The judge rule from the bench and didn't say much (though we he did say wasn't great -- now same as a well considered written opinion though). If interested, there is some good discussion following my LinkedIn post linked below. Really love your hallucination tracking project too! (Used it as a jumping off point on why lawyers should use LLMs for legal research anyway.... sometimes, and carefully).
https://www.linkedin.com/posts/activity-7427218654511173632-0vJM?utm_source=share&utm_medium=member_desktop&rcm=ACoAABW5pB0BRDTgxtNoNOPeS2-G194yZ_rLSb4
Ah ! That's a big correction in fact, I totally missed it, will fix !
Not sure if you saw the comments, but here more for the Gell-Mann amnesia bit. Someone asked Schumer about AI legal hallucinations and his response was “Use the most recent models (try GPT-5.2 Pro in ChatGPT, GPT-5.3-Codex in Codex, Opus 4.5 in Claude) and you won't have this issue.” Hard to imagine someone is that unaware of the limits of LLMs.
https://x.com/mattshumer_/status/2021369465284468955?s=46&t=g0YG1OIX7LFM0CY5ZK_Plg
Yes, I saw that too ! That's part of the broader credibility point I am making in the footnote, he did not strike me as anyone worth listening to. But he hit a nerve and went viral nonetheless.