AI Circle Spat Over Using ChatGPT as Legal Authority in Wentworth Funding Row
Amid the dispute over Dustin Moskovitz withdrawing funding from Wentworth in AI safety circles, jdpressman and ohabryka clashed repeatedly over whether publicly making statements that could enter litigation discovery constitutes a funding risk. Both sides cited ChatGPT's legal analysis as support, raising questions about the reliability of treating LLM answers as legal grounds. The two remain at odds with no consensus, but the spat exposes how heavily the AI community leans on LLM output for legal risk judgment.
Confirmed
- jdpressman wrote a long comment arguing one reason ChatGPT can't replace a lawyer is that it's trained to avoid harshly condemning anyone's behavior, and those training biases shape its legal conclusions.
- jdpressman stated his position: from a Bayesian view, someone who can't resist making public statements that could trigger discovery while under intense public scrutiny is someone he wouldn't fund—and that judgment has nothing to do with ChatGPT's take on which specific acts are illegal.
- Responding to ohabryka's criticism, jdpressman admitted his phrasing was "vague" but stood by his core legal judgment: if plaintiffs' lawyers believed he witnessed relevant events, his DMs with ohabryka could indeed enter litigation discovery; he also clarified he isn't in SF and clearly witnessed nothing.
- ohabryka pushed back against jdpressman and Richard Ngo, claiming "this is not a substantive legal risk, go ask your LLMs," and questioned why everyone's views on legal threats in this area are both confident and foolish.
Why it matters
- The dispute shows a real case of LLM output being cited as legal authority, while both sides simultaneously question the reliability of each other's LLM answers—highlighting the lack of consensus and professional vetting in how the AI community uses legal tooling
- The underlying controversy (the funding pull and public-statement risk) touches the trust boundary between funders and grantees, and may shape how people in the circle self-censor future public statements
2026-10-13 ~ 2026-10-13 · 5 related posts
Primary sources
- [source] 'Ask your LLMs': legal-risk spat erupts in the Wentworth funding thread — ohabryka · 2026-10-13
- [source] Using ChatGPT as a lawyer in the Dustin-Wentworth funding dispute — jd_pressman · 2026-10-13
- Bayesian take on the Wentworth dispute: public statements as funding risk — jd_pressman · 2026-10-13
- [source] AI Twitter Drama: Can You Trust ChatGPT Answers About Legal Discovery Risk? — ohabryka · 2026-10-13
- jd_pressman Doubles Down: My DMs Could Be Discoverable If Lawyers Believed I Witnessed It — jd_pressman · 2026-10-13