Brazil's Courts Aren't Waiting for AI Regulation

Chatbot hallucinations are no longer the primary concern surrounding artificial intelligence. They still spew false information, but more concerning, autonomous AI agents are now making decisions and acting on their own. They’ve even hacked into private computer networks.

While the US Congress continues to debate what, if anything, to do about the risks posed by AI, the Brazilian judiciary has already acted. In 2025, Brazil’s National Council of Justice (CNJ) adopted rules governing how artificial intelligence can be developed and used across its court system.

Resolution 615/2025 regulates the development, use, and governance of generative AI throughout the Brazilian judiciary. Among other things, it requires AI systems to be transparent and explainable. It also requires human oversight, particularly when AI is used in ways that could affect judicial decisions.

The rules classify AI systems based on their level of risk. Low-risk uses include routine administrative tasks, such as document formatting and file organization. High-risk systems are those that could influence legal outcomes or affect individual rights. These include tools that help evaluate evidence or draft judicial decisions.

The closer AI gets to influencing a legal outcome, the tighter the controls become. High-risk systems face additional risk assessments, monitoring, and audits. Some uses are outright prohibited, including systems that prevent meaningful human review or leave judges completely dependent on an AI-generated result.

Brazil’s approach isn’t to keep AI out of the courtroom. It’s to make sure AI never replaces the person responsible for the decision. As AI becomes more autonomous, that distinction becomes increasingly important.