Court document: a Colombian judge used ChatGPT to ask legal questions about a case and included its responses in his decision, to save time writing the judgment
Janus Rose / VICE :
Context & Ripple Effects
This report captured the first widely covered case of a sitting judge putting ChatGPT output directly into a court decision — not as background research but quoted in the judgment itself, explicitly to save drafting time. It landed months before any judicial institution had issued rules on the practice.
What came after shows the pattern this single case kicked off: within months, a New York lawyer faced sanctions over a ChatGPT-written brief full of hallucinated citations, and by year's end the UK Judicial Office issued guidance permitting judges to use AI for rulings while flagging its shortcomings. The Colombian judge's improvisation has since become the baseline behavior that regulators are racing to formalize.
First-order effects
- The judge himself bears immediate exposure: his published decision now contains unverified model answers on legal questions, making him personally accountable for any errors the model introduced.
- OpenAI's tool gained its first documented role inside a state's judicial process without any institutional review of whether its legal answers were accurate.
Second-order effects
- Lawyers' hallucinated-citation scandals forced the issue onto regulators' agendas — sanctions cases like the New York brief turned informal judge experimentation into a compliance problem that bar authorities and judiciaries had to answer with explicit policies.
- Courts under backlog pressure elsewhere followed the same logic: US judges are now openly testing generative AI to expedite legal research, summarize cases, and draft routine orders.
Third-order effects
- If adoption keeps outpacing verification, the judiciary faces a structural auditability gap — decisions whose reasoning chains include machine-generated text that no one independently checked — pushing institutions toward codified disclosure rules rather than bans.
- In Latin America specifically, courts adopting AI tools to clear backlogs while simultaneously struggling to adjudicate AI-generated evidence means the same institutions become both heavy users and arbiters of the technology.
The trend: Judicial systems are moving from improvised individual use of generative AI toward codified, disclosed usage rules, driven by backlog pressure and punctuated by hallucination scandals.