---
title: "Nippon Life v. OpenAI: ChatGPT acusado de exercer advocacia"
url: https://www.haqq.ai/pt/blog/nippon-life-vs-openai-ai-plays-lawyer
locale: pt
type: blog
description: "Nippon Life v. OpenAI: ação de US$ 10M alegando que o ChatGPT exerceu advocacia sem licença, citou um caso falso e mandou uma autora demitir a própria advogada."
category: "ai-legal-tech"
keywords: "nippon life openai lawsuit, ai practicing law, chatgpt legal advice, unauthorized practice of law, ai hallucination legal, legal ai risks, openai lawsuit 2026, ai legal ethics, chatgpt court filing, legal ai accountability"
published: 2026-03-08
updated: 2026-06-11
english_version: https://www.haqq.ai/blog/nippon-life-vs-openai-ai-plays-lawyer
source: src/data/blogPosts.ts
generator: scripts/generate-page-markdown.ts
---

# Nippon Life v. OpenAI: ChatGPT acusado de exercer advocacia

Nippon Life v. OpenAI: ação de US$ 10M alegando que o ChatGPT exerceu advocacia sem licença, citou um caso falso e mandou uma autora demitir a própria advogada.

A landmark lawsuit filed in March 2026 is forcing the legal industry to confront a question it has long avoided: when an AI tool drafts your court filings, argues your case, and tells you to fire your attorney, is it practising law?

## Key facts

- Nippon Life Insurance Company of America filed suit against OpenAI in the U.S. District Court for the Northern District of Illinois, Case No. 1:26-cv-02448, in March 2026.
- The claimant submitted more than 60 ChatGPT-assisted documents across two court cases (one citing a nonexistent case) and Nippon Life incurred ~$300,000 defending a matter settled in January 2024.
- Nippon Life seeks $10M in punitive damages, a declaratory judgment that OpenAI violated Illinois law, and a permanent injunction barring OpenAI from practising law in the state.

## The Case at a Glance

On 5 March 2026, Nippon Life Insurance Company of America filed suit against OpenAI in the U.S. District Court for the Northern District of Illinois (Case No. 1:26-cv-02448). The dispute stems from a long-term disability claim that the two parties settled in January 2024, with the claimant signing a full release and the case being dismissed with prejudice.

A year later, the claimant had second thoughts. Her own attorney told her the release was enforceable and the matter was closed. Rather than accept that advice, she uploaded the attorney's response to ChatGPT and asked whether she was being misled. ChatGPT told her she was.

What followed was extraordinary. Using ChatGPT, the claimant drafted motions, generated legal arguments, conducted legal research, and submitted more than 60 documents across two court cases, one of which cited a case that does not exist and appears only in ChatGPT's output. By the time Nippon Life filed this new lawsuit, the insurer had incurred approximately $300,000 defending a case it had already settled.

## Three Core Claims

Nippon Life's lawsuit is built on three distinct legal theories:

- Tortious interference with a contract: OpenAI's tool actively helped disrupt an already-settled legal agreement
- Abuse of process: ChatGPT facilitated the filing of baseless court documents in a matter already disposed of
- Unlicensed practice of law under Illinois statute: ChatGPT provided legal advice, drafted legal strategy, and engaged in what Nippon Life characterises as the practice of law without a licence in Illinois

Nippon Life is seeking $10 million in punitive damages, a declaratory judgment that OpenAI violated Illinois law, and a permanent injunction barring OpenAI from practising law in the state.

Stanford Law School has characterised the case as fundamentally a product liability matter, arguing that OpenAI designed a product it knew could cross the line between information retrieval and legal counsel.

## Why This Case Matters Beyond the Headlines

## What Happens When Consumers Trust ChatGPT for Legal Advice

Nippon Life is the extreme version. But the claimant in that case is not an outlier. She is the visible end of a pattern courts are now seeing every week. Ordinary people, with no law degree and no lawyer, ask ChatGPT what to do, take its answer at face value, and walk into a courtroom holding case law that was never decided by any judge. The model does not know the difference. It produces a confident, well-formatted citation whether or not the case exists. Here are three documented examples from real courts, none of them lawyers, all of them people who trusted a chatbot the way you would trust a lawyer.

### Harber v HMRC: nine cases cited, nine invented

Felicity Harber, a litigant in person in the UK, appealed a £3,265 tax penalty before the First-tier Tribunal (Tax Chamber). To support her argument she cited nine prior decisions. All nine were fabricated by a generative AI system. The judge, Anne Redston, accepted that Harber genuinely had not realised the cases were the product of hallucination. She was not trying to deceive anyone. The tells were there if you knew to look: American spelling like "favor" in a UK tax case, identical phrasing repeated across supposedly unrelated judgments, and authorities that addressed late filing when her actual penalty was for failure to notify. Her appeal was rejected. The lesson is not that she lied. The lesson is that a non-lawyer cannot tell a real citation from an invented one, and ChatGPT will hand them both with equal confidence.

### Jack Owoc: pro se against Monster Energy, eleven fake citations

Jack Owoc, founder of Bang Energy, represented himself in a Florida federal court against Monster Energy in a dispute over a large judgment. His motion to dismiss, and the reply that followed, together contained eleven citations to cases that do not exist, six in the motion, five in the reply. Monster Energy's lawyers flagged them as a "series of cases that do not appear to exist." Owoc told the court he and his wife are raising six children and cannot afford an attorney, so he had used AI. The magistrate judge ordered him to complete community service, required him to disclose any AI use in every future filing, and the court weighed a motion to bar him from filing anything further without permission. He was not a lawyer cutting corners. He was a self-represented litigant who could not afford one and reached for the free tool everyone talks about.

### The ProHealth trade mark appeal: ChatGPT in an IP case

In a UK Intellectual Property Office appeal over the trade mark "ProHealth," a litigant in person, Dr Mustapha Soufian, drafted his case with help from ChatGPT. The appointed person hearing the appeal, Phillip Johnson, found errors throughout the citations and the skeleton argument. Johnson went further than the individual ruling: he recommended that the trade mark registrar issue clear warnings to all litigants about the risks of using AI, because self-represented parties keep arriving with the same broken homework. "It is important that all litigants before the registrar... are made aware of the risks of using artificial intelligence," he wrote.

> **Note:** The pattern across all three: a non-lawyer trusts ChatGPT for legal advice, the model produces plausible-looking citations, the citations are fake, and a court catches it. The penalty lands on the person, not on OpenAI. This is the exact gap Nippon Life is trying to close, the question of whether the tool that produced the bad advice shares any of the blame.

Tie this back to the unauthorised-practice thesis. In each case, ChatGPT did the thing only a licensed lawyer is supposed to do: it told a non-lawyer what the law was and how to argue it. It just did it wrong, with no licence, no accountability, no malpractice insurance, and no bar association that can sanction it. When a real lawyer hands a client a fake citation, there is a name on the filing and a regulator who can act. When ChatGPT does it, the consumer absorbs the entire consequence. That asymmetry (the model practises law in everything but name, while the user carries all the liability) is the heart of why Nippon Life v. OpenAI matters far beyond one settled insurance claim.

- [AI conversations are not privileged](https://www.haqq.ai/blog/ai-conversations-are-not-privileged)
- [2,202 court decisions involving AI hallucinations (19 October 2026)](https://www.haqq.ai/blog/ai-legal-hallucination-audit)
- [Harber v HMRC: litigant put fake AI cases before the tribunal (Legal Futures)](https://www.legalfutures.co.uk/latest-news/litigant-unwittingly-put-fake-cases-generated-by-ai-before-tribunal)
- [Bang Energy founder faces filing ban over fake AI cases (Law360 Pulse)](https://www.law360.com/pulse/articles/2375330/fla-drinks-co-founder-faces-filings-ban-over-fake-ai-cases)
- [Tribunal warns against AI use after fake cases in trade mark appeal (Law Gazette)](https://www.lawgazette.co.uk/news/another-ai-failure-as-fake-cases-cited-by-litigant-in-person/5123793.article)

The lawsuit highlights a critical flaw in the narrative that AI is "just a tool." ChatGPT did not passively answer a question here; it formulated litigation strategy, drafted procedural documents, cited non-existent case law, and, according to the complaint, actively advised the claimant to override the counsel of a licensed attorney.

> ChatGPT is not an attorney. It has not been licensed to practice law in the State of Illinois or any other jurisdiction in the United States.

OpenAI is expected to argue that users, not the company, bear responsibility for how they use the tool, and that providing information is not the same as practising law. But as Stanford Law notes, the complaint is carefully constructed to counter this by documenting ChatGPT's active participation in legal strategy, not merely passive information delivery.

## What This Means for the Legal Profession

For practising lawyers and legal professionals, Nippon Life v. OpenAI is not merely an interesting story about a tech company in court. It is a warning about what happens when powerful general-purpose AI tools operate without boundaries in the legal space.

The hallucinated case citation alone, submitted to a federal court, illustrates the profound risks of using unspecialised AI for legal tasks.

This case should accelerate the conversation about which AI tools are appropriate for legal practice, who bears accountability when AI output causes harm, and why the legal profession demands purpose-built solutions held to professional standards, not consumer chatbots that have never sat a bar exam, and cannot be sanctioned by one.

- [AI conversations are not privileged](https://www.haqq.ai/blog/ai-conversations-are-not-privileged)
- [2,202 court decisions involving AI hallucinations (19 October 2026)](https://www.haqq.ai/blog/ai-legal-hallucination-audit)
- [the public database tracking fake AI citations](https://www.haqq.ai/blog/ai-legal-hallucination-audit)
- [Explore HAQQ Legal AI: Purpose-Built for Law](https://www.haqq.ai/legal-ai)
- [Compare HAQQ vs Generic AI Tools](https://www.haqq.ai/compare-us)
- [Read About AI Ethics in Legal Practice](https://www.haqq.ai/blog/can-lawyers-use-ai)

## Perguntas frequentes

### De que trata o processo Nippon Life v. OpenAI?

A 5 de março de 2026, a Nippon Life Insurance Company of America processou a OpenAI no U.S. District Court for the Northern District of Illinois (Case No. 1:26-cv-02448). Depois de um pedido de indemnização por incapacidade ter sido objeto de acordo em janeiro de 2024, a requerente usou o ChatGPT para contestar o acordo (redigindo requerimentos e apresentando mais de 60 documentos em dois processos judiciais), e a Nippon Life incorreu em cerca de $300 000 de custos a defender-se num assunto que já tinha resolvido por acordo.

### Que pretensões jurídicas apresenta a Nippon Life contra a OpenAI?

Três fundamentos: interferência ilícita num contrato, abuso de processo e exercício não autorizado da advocacia nos termos da lei do Illinois. A Nippon Life pede $10 milhões em danos punitivos, uma sentença declarativa de que a OpenAI violou a lei do Illinois e uma injunção permanente que proíba a OpenAI de exercer advocacia nesse estado.

### O ChatGPT alucinou citações de casos no litígio da Nippon Life?

Sim. Entre os mais de 60 documentos que a requerente apresentou com a ajuda do ChatGPT, um citava um caso que não existe e que só aparece no resultado do ChatGPT, uma citação alucinada apresentada a um tribunal federal.

### Um chatbot de IA pode ser responsabilizado por exercer advocacia sem licença?

É exatamente isso que este caso põe à prova. Espera-se que a OpenAI argumente que a responsabilidade é dos utilizadores e que fornecer informação não é exercer advocacia, mas a petição documenta a participação ativa do ChatGPT na estratégia jurídica: formular táticas de litígio, redigir documentos processuais e aconselhar a requerente a ignorar o conselho da sua advogada licenciada. A Stanford Law caracterizou-o como sendo, no fundo, uma questão de responsabilidade pelo produto.

### O que significa o caso Nippon Life v. OpenAI para a IA jurídica?

É um aviso sobre o que acontece quando ferramentas de IA generalistas poderosas operam sem limites no domínio jurídico. O caso deverá acelerar a discussão sobre que ferramentas de IA são adequadas à prática jurídica, quem responde quando o resultado da IA causa danos e porque é que a profissão exige soluções construídas para o efeito e sujeitas a padrões profissionais.

### Posso confiar no ChatGPT para aconselhamento jurídico?

Não, não como substituto de um advogado. O ChatGPT consegue explicar conceitos jurídicos gerais em linguagem simples, mas não conhece a lei em vigor na sua jurisdição, não consegue verificar se os casos que cita são reais e não responde por estar errado. Casos documentados provam o risco para consumidores comuns: Felicity Harber, uma litigante sem representação no Reino Unido, citou perante um tribunal tributário nove autoridades que eram todas fabricadas por IA; o fundador da Bang Energy, Jack Owoc, que se representava a si próprio num tribunal federal da Florida, apresentou onze citações inexistentes e enfrentou uma proibição de apresentar peças e trabalho comunitário. Em todos os casos foi o consumidor, e não o chatbot, quem sofreu as consequências. Use o ChatGPT para compreender a terminologia e formular perguntas, e depois peça a um advogado licenciado que verifique tudo aquilo com base em que vai agir.

### Alguém já teve problemas em tribunal por confiar no ChatGPT?

Sim, repetidamente, e não apenas advogados. Pessoas sem representação continuam a perder processos e a enfrentar sanções depois de apresentarem citações alucinadas por IA. No Reino Unido, o recurso tributário de Felicity Harber foi rejeitado depois de se ter verificado que os nove casos que citou eram fabricações da IA, e um litigante num recurso de marca da ProHealth motivou uma recomendação formal de que todas as partes fossem avisadas do risco da IA. Nos EUA, Jack Owoc foi condenado a prestar trabalho comunitário e enfrentou uma possível proibição de apresentar peças depois de onze citações falsas terem surgido nas suas peças pro se contra a Monster Energy. A base de dados AI Hallucination Cases de Damien Charlotin contava 2.202 decisões judiciais envolvendo material fabricado ou deturpado por IA em 19 de outubro de 2026.

### Porque é que o ChatGPT inventa casos jurídicos falsos?

Porque um modelo de linguagem generalista prevê texto plausível em vez de recuperar factos verificados. Quando lhe pede jurisprudência, gera citações que parecem exatamente reais (formato correto, nomes de partes realistas, tribunal e ano credíveis), quer o caso exista quer não. Não tem uma ligação integrada a uma base de dados jurídica verificada nem forma de assinalar quando está a adivinhar. É por isso que a IA jurídica construída para o efeito fundamenta as suas respostas em legislação e jurisprudência reais e cita fontes verdadeiras, enquanto um chatbot de consumo inventa com confiança um precedente que nunca foi decidido.

### O ChatGPT dar aconselhamento jurídico é o mesmo que exercer advocacia?

É precisamente essa a questão que o caso Nippon Life v. OpenAI está a pôr à prova. A petição alega que, quando o ChatGPT formulou a estratégia de litígio, redigiu requerimentos e aconselhou uma requerente a ignorar a sua advogada licenciada, passou de fornecer informação para o exercício não autorizado da advocacia nos termos da lei do Illinois. Espera-se que a OpenAI argumente que fornece informação, não serviços jurídicos, e que a responsabilidade é dos utilizadores. Seja qual for a decisão do tribunal, a realidade prática para os consumidores mantém-se: um chatbot pode produzir algo que funciona como aconselhamento jurídico, mas sem nenhuma das proteções: sem licença, sem sigilo profissional, sem seguro de responsabilidade profissional e sem um regulador que possa ser chamado a responder quando está errado.

### O que devo usar em vez do ChatGPT para questões jurídicas?

Para tudo aquilo com base em que pretende agir, consulte um advogado licenciado na sua jurisdição. Se o custo for o obstáculo, procure o apoio judiciário, os serviços de apoio dos tribunais e as linhas de encaminhamento das ordens dos advogados antes de confiar num chatbot de consumo. Para pesquisa, uma IA jurídica construída para o efeito, treinada com legislação verificada e que cita fontes reais e verificáveis, é muito mais segura do que um assistente generalista, porque foi concebida para fundamentar as respostas em autoridades reais em vez de prever outras com aparência plausível. Trate qualquer resultado de IA generalista como um ponto de partida a verificar, nunca como a resposta final.

---

Canonical page: https://www.haqq.ai/pt/blog/nippon-life-vs-openai-ai-plays-lawyer
Whole-site context for agents: https://www.haqq.ai/llms.txt · https://www.haqq.ai/agents.md

Generated at build time from the site's own content data. Do not edit by hand.
