AIAcademy · AIAcademy · 2026-05-16
Read the AI Incident Database round-up
Damien Charlotin's globally-maintained register now logs 1,227 court filings containing AI-fabricated citations, with new cases catalogued weekly through Q1 2026. The number itself is striking, but the structure of the curve is the actual lesson — this is the cleanest case study in the field of what "AI literacy as a professional competency" looks like when courts get tired of explaining it.
The arc runs from Mata v. Avianca (June 2023, the original Steven Schwartz / ChatGPT fake-citation sanction, ~$5,000) through Morgan & Morgan in 2025 (the largest US personal-injury firm; drafting attorney fined $3,000 with temporary bar admission revoked) to Judge Mark D. Clarke's $110,000 sanction against two Oregon lawyers in April 2026 — the costliest single AI-hallucination penalty in US legal history. The Sixth Circuit added $30,000 in March 2026 for 24+ fake citations in a single brief. Q1 2026 alone produced >$145,000 in cumulative US-court sanctions. Nebraska handed down the first license suspension in early 2026. The Third Circuit issued a panel reprimand for seven fabricated citations plus a non-existent case.
What changed is not the technology. It is the assumed standard of care. Florida Bar Ethics Opinion 24-1, Miami-Dade's AI Disclosure Order (Feb 2026), and a string of state-supreme-court directives now treat verification of every AI-generated citation as mandatory before filing. "I trusted the tool" is no longer a defence; it is the violation. Several bar associations have moved faster than legislatures here — the discipline regime is catching up to the technology before the statutes are.