The Dangers of AI: The Hallucination Tragedy
When the pursuit of automation becomes illegal — how AI-fabricated legal citations turned into six-figure court sanctions.
Co-written with Carlos Alvarez for Axiom Pathways.
Traces the legal profession’s AI hallucination problem from a 2023 case where a lawyer submitted six ChatGPT-fabricated court citations, through a 2026 Oregon ruling that fined two attorneys $110,000 for 23 fake citations in one brief, to the more than 1,200 documented cases of fabricated legal citations tracked by a HEC Paris researcher. The piece argues the story isn’t simply “AI is dangerous” — adoption among lawyers hit 83% because AI-assisted contract review is genuinely faster and more accurate than human review — but that the real disruption is structural: AI is quietly eliminating the entry-level document-review work junior associates have historically used to learn the profession, forcing firms to redesign how new lawyers are trained.