Use of AI in legal documents: responsibility remains with the lawyer.

Brazilian courts at different levels—the Superior Court of Justice (STJ), the Superior Labor Court (TST), and Regional Courts—have been issuing rulings concerning the use of generative artificial intelligence in drafting petitions and appeals. The common thread among these judgments is clear: the responsibility for the veracity of the content presented in court remains entirely with the signing attorney, regardless of the tool used to produce it.

In May 2026, the Superior Court of Justice (STJ), in a petition addressed to the Court's Vice-Presidency, identified hidden commands designed to induce artificial intelligence systems to form artificially favorable conclusions for the appeal—an episode treated as a violation of the duty of procedural good faith. Cases of citing non-existent jurisprudence or doctrine, generated by AI without human review, have also resulted in convictions for bad-faith litigation and notification to the Brazilian Bar Association (OAB).

What the courts have decided

• The use of AI is not, in itself, grounds for sanction — the problem lies in the absence of human review before the document is submitted.
• Fictitious or falsified case law citations have led to fines for bad-faith litigation and, in more serious cases, to the dismissal of the case without a decision on the merits.
• Responsibility does not transfer to the tool or to third parties: the duty of verification lies with the professional who signs the document.

Recommendation

Law firms and legal departments that use artificial intelligence in drafting legal documents must maintain a mandatory step of human review and verification of all citations, summaries, or precedents before filing. It is recommended to document this verification process as part of routine procedural compliance. We continue to monitor case law on this topic and are available to guide the responsible adoption of these tools.

Article written by: Rafael Mello.