AI Legislation in Brazil Follows the European Model

Bill 2.338/22, which regulates Artificial Intelligence in Brazil, is entirely based on the European Union’s AI Act and is founded on the principle that the greater the risk to fundamental rights, the greater the responsibility placed on developers, explained Appellate Court Judge Demócrito Reinaldo Filho of the Pernambuco State Court of Justice (TJ-PE). He spoke during the final plenary session of the 44th ABPI Congress on Tuesday, August 13th, alongside ABPI President Gabriel Leonardos and the organization’s Vice President, Tatiana Campello.

According to the Appellate Court Judge, who has deeply studied the subject, the AI Act—like Bill 2.338/22—categorizes the risks associated with Generative AI systems into three categories. In cases of unacceptable risk, such as those affecting individual health or threatening institutions, the system is eliminated. He cited China’s social credit scoring system as an example, which infringes on fundamental rights. “Some systems that pose unacceptable risks to fundamental rights and democratic institutions must be banned and cannot be offered in the market,” the Judge added.

In the next category, which involves a lower but significant level of risk, the legislation remains stringent in holding developers accountable. The third category, involving minimal risk to individuals and institutions, imposes fewer legal obligations, primarily on transparency. Another key aspect of the AI Act, the Judge noted, is the requirement for developers to comply with the European copyright directive. “It is crucial to regulate general-purpose AI systems to establish accountability for each participant in the AI product’s value chain,” he said.