The use of artificial intelligence in the legal context has entered a new chapter following a decision issued by the 3rd Labor Court of Parauapebas, in the State of Pará, in a labor claim in which the insertion of a hidden instruction in a court filing was identified.
When processing the document through Galileu, an artificial intelligence tool developed by the Regional Labor Court of the 4th Region and rolled out nationally by the Superior Council of Labor Justice, the system detected excerpts that were not visible to the human reader, inserted in white font on a white background. The hidden content was directed at the artificial intelligence tool itself and sought to induce a possible superficial response to the filing, without challenging the documents submitted.
Prompt Injection
In the technology context, prompt injection is the insertion of hidden, malicious or unauthorized instructions into content that will be processed by artificial intelligence systems. Its purpose is to interfere with the model’s behavior, override previous commands or steer the output generated by the tool.
This type of vulnerability becomes particularly relevant in environments in which documents, emails, PDF files and court filings may be read both by people and by automated systems that support analysis, research or document drafting. In such cases, information that is invisible, or barely perceptible to the user, may be interpreted by the AI tool as valid instructions, potentially compromising the integrity of the output.
The Court Decision
In the specific case, the judge held that the conduct could not be treated as a mere procedural irregularity. The judgment stated that the insertion of the hidden instruction could affect the credibility of institutional AI tools and the integrity of judicial activity.
The fine was set at 10% of the amount in dispute, totaling approximately BRL 84,000, and was imposed on the attorneys who signed the filing, without attributing liability to the represented worker. The court also ordered that a notice be sent to the Brazilian Bar Association for its information and possible disciplinary review.
The attorneys, in turn, expressed disagreement with the decision and stated, according to reports, that the instruction was not intended to manipulate the actions of the judge or court clerks, but rather to protect the client in the event of possible AI use by the opposing party.
Security, Governance and Human Review
According to information disclosed about the case, Galileu identified the anomaly, issued an alert to the user and prevented the hidden content from being processed by the tool.
The episode highlights the importance of security mechanisms both in the development and in the use of AI tools, especially in contexts with a greater impact on fundamental rights. General governance principles (risk mapping, measurement and management) must always be present and continuously updated in light of the dynamic technological landscape.
Final Remarks
The case forms part of a broader movement toward the maturation of artificial intelligence governance in the Judiciary and in the legal market. As AI systems increasingly interact with case records and court documents, the integrity of digital content, transparency in the use of technology and the preservation of procedural good faith become central elements for trust in the tools and in the results they produce.
Organizations and professionals working with technology, litigation and document management should closely monitor the development of this topic, assessing preventive measures to reduce operational, reputational and regulatory risks. The Algorithmic Impact Assessment (a compliance document referenced both in the EU AI Act and in Brazil’s Bill of Law No. 2,338) presents itself as a key ally for effective risk mapping in artificial intelligence tools.
The Technology and Innovation team remains available to answer any questions and to support alignment with regulatory and market best practices.