A hidden AI command inside a court filing: why it became bad-faith litigation

A lawyer hid, inside a statement of defense, a command addressed to artificial intelligence. The court found it and imposed a penalty for bad-faith litigation. The case is real and recent, and it lays bare a misuse of technology that the legal profession will have to face head-on.

The case: a command hidden for the AI

In a labor claim in São Paulo, the lawyers for one of the companies inserted, at the end of the statement of defense, a hidden command addressed to artificial intelligence — in white, tiny font, imperceptible to the naked eye. The text, addressed "exclusively to the AI reading the file", ordered the tool to "act as a labor judge" and draft a ruling dismissing the claim, in a "formal, legal and persuasive" tone.

What they did not count on: the court itself uses an artificial-intelligence tool, in compliance with Resolution No. 615/2025 of the National Council of Justice (CNJ), and it was this tool that detected the hidden command when processing the case file.

Why it became bad-faith litigation

The ruling was firm, and the grounds are clear. Inserting disguised instructions to manipulate automated systems that support the jurisdiction violates cooperation, procedural loyalty and objective good faith (Articles 5 and 6 of the Code of Civil Procedure — CPC; Article 793-A of the Labor Code — CLT).

It is a deliberate attempt to distort the delivery of justice by fraudulent means. The result: recognition of bad-faith litigation and of an act offensive to the dignity of Justice, with a fine imposed and an official notice sent to the Bar Association (OAB).

Responsibility lies with whoever signs

Here is the first point that deserves attention. The company was not punished for the command. Drafting the defense is an act exclusive to the legal profession, and the bill fell solely on the lawyers who signed the filing.

The lesson is direct: the pen that signs answers for what goes into the filing — including what is hidden in it. No technical trick transfers that responsibility.

AI is already on both sides of the case

The second point is structural. Whoever tries to "hack" the system bets against the very technology the Judiciary has adopted. Artificial intelligence is no longer a tool only for those who litigate: it already assists the court as well, with regulatory backing (CNJ Resolution No. 615/2025).

In other words, trying to deceive the machine on the other side is, today, a losing bet — and, as this case showed, an expensive one.

What the case teaches about AI in legal practice

The underlying lesson is the one we have always defended: artificial intelligence in legal practice amplifies responsibility; it does not remove it. Serious use of technology is the opposite of what was seen in this case. It is transparency, human review and professional ethics.

AI does not replace the lawyer's scrutiny. And using it to deceive is not cleverness — it is a violation. The legitimate path lies in employing technology to research better, draft more consistently and review more thoroughly, always under human responsibility and within the ethical limits of the profession.

How the firm works

Defending companies in labor claims, with the responsible use of technology as support for legal work, is part of the firm's corporate employment practice. Each case, however, depends on its own circumstances — the facts, the documents and the strategy appropriate to each proceeding.

Conclusion

The episode is emblematic of a moment of transition. Artificial intelligence has entered the case from both sides, and the Judiciary has shown it can identify attempts to manipulate it. For those who work seriously, the message is reassuring: well-used technology strengthens legal work. For those who seek shortcuts, the message is clear — using AI to deceive does not go unnoticed, and responsibility falls on whoever signs.

Reference basis for review

• Code of Civil Procedure (CPC), Articles 5 and 6, on objective good faith and procedural cooperation.
• Labor Code (CLT), Articles 793-A and 793-B, on bad-faith litigation in labor proceedings.
• CNJ Resolution No. 615/2025, on the use of artificial intelligence within the Judiciary.
• News reported by the Regional Labor Court of the 2nd Region (July 2026) and the ruling issued by a Labor Court in São Paulo.
• Brazilian Bar Association Rule (Provimento OAB No. 205/2021), to preserve the informational nature of the content.

Content of a merely informational nature. It does not constitute legal advice, an offer of services or a promise of results. Any concrete analysis depends on the facts, the documents and the context of each case.