Sanction teardown · TJPR, 13ª CC, Brazil · 2026-04-08
Agravo de Instrumento Nº 0104033-49.2025.8.16.0000
What happened
In TJPR, 13ª CC, Brazil, a filing relied on an unnamed/unconfirmed AI tool to help draft legal argument. The court identified the following problems with the citations in that filing:
-
Misrepresented (Case Law)Cited 'STJ, AgInt no AREsp 1368223/SP, Rel. Min. Benedito Gonçalves, DJe 28/03/2019' — court found the referenced decision was mischaracterized (actually a 2016 matter on criminal procedure) and not pertinent to the argued issue.
-
Fabricated (Case Law)Alleged 'TJPR, AI 0033665-93.2022.8.16.0000, Rel. Des. Guilherme Frederico Hernandes Denz, J. 24/08/2023' was asserted by counsel but court's search showed this Agravo de Instrumento was invented and does not exist.
-
Fabricated (Case Law)Citation 'STJ, RESP1984567 / PR, Rel. Min. LUIS FELIPE SALOMÃO, julgado de 5.4.22' was indicated by counsel but court concluded the referenced precedent was inexistent or improperly alleged.
Which AI tool
an unnamed/unconfirmed AI tool. Note: Charlotin's public database records tool attribution only where a court order, brief, or reporting on the matter states it explicitly; "unidentified" or "implied" means the record indicates AI use but does not name a specific product — we do not guess.
Outcome
Court found the use of nonexistent or mischaracterized jurisprudence constituted litigância de má-fé; imposed a fine of 1% of the updated value of the cause and ordered communication to OAB/PR for possible disciplinary follow-up. (monetary penalty: 1 .)
Additional detail
The court determined that the exequente (through counsel) submitted jurisprudence that was nonexistent or mischaracterized — apparently produced by AI — to support its positions. The panel held this behavior amounted to litigância de má-fé (art. 80 II CPC), reduced the pecuniary sanction to 1% of the updated value of the cause (R$1,589,690.46 → R$15,896.90), and ordered communication to OAB/PR. The tribunal emphasized that counsel must verify AI outputs and that citing fabricated or altered precedents can induce the court to error and exceed excusable research mistakes.
How Citation Safe would have caught this
Citation Safe runs three deterministic layers before a brief is filed: (1) does the citation exist against CourtListener's database of published opinions, (2) if quoted, does that exact language appear in the source, (3) does the cited case actually support the proposition it is cited for. Fabricated case citations fail Layer 1. Fabricated or misattributed quotations fail Layer 2 even when the underlying case is real. Misrepresented holdings — a real case cited for a proposition it does not support — are the target of Layer 3. None of these checks involve asking another language model whether the citation looks right; they are lookups and text-matches against the actual source, which is why a hallucinated citation has to survive a direct lookup against the authoritative source — not another model's opinion — to earn a VERIFIED stamp; our measured false-verify rate is published live at /quality.
Check a brief before you file it → · See our live false-verify rate
Source: https://www.damiencharlotin.com/documents/2884/0104033-49.2025.8.16.0000.pdf, via Damien Charlotin's public AI Hallucination Cases Database (CC0).