A court in São Paulo has temporarily barred Meta from using automated tools to collect news reports, photographs, videos and other material published by media outlets represented by three Brazilian industry associations. The company is also prohibited from using that content to develop or train artificial intelligence systems, including Meta AI.
The ruling, issued Tuesday, November 3, followed a joint action by the National Newspaper Association (ANJ), the Brazilian Association of Radio and Television Broadcasters (Abert) and the National Association of Magazine Publishers (Aner). The groups sought to suspend new Meta commercial terms scheduled to take effect on November 20.
The associations said the rules would allow Meta to use automated programs to access, copy, store and analyze content on publishers’ websites. The court suspended provisions requiring businesses using Meta’s platforms to authorize the use of web scrapers and other tools to collect material from domains and URLs connected to their services, without compensation. Under the terms, continued use of the platforms would count as acceptance.
Meta has five days to respond to the urgent request. The associations have 30 business days to file their main claim, in which the merits of the case will be considered. A breach of the order carries a fine of 1 million reais per day, capped at 100 million reais. The court also ordered Meta to keep the accounts, pages and tools of the associated outlets active, without restrictions or penalties for rejecting the new terms.
Judge Guilherme Madeira Dezem of São Paulo’s 44th Civil Court said the penalty reflected Meta’s alleged reluctance to comply with rulings in other cases. He also found that scraping news websites to supply language models could constitute reproduction of copyrighted works, which Brazilian law requires authors to authorize in advance and expressly. The judge said the country’s legislation does not provide an exception allowing text and data mining to train commercial AI systems.
The ruling noted that Meta’s first version of the changes, announced on October 16, explicitly referred to training systems. Although the company later removed that wording following criticism from industry groups, the judge said the mechanisms for large-scale content extraction remained in place.
Dezem also said the terms could force publishers to claim rights over material they do not wholly own, such as agency photographs, illustrations and contributors’ work. That could leave outlets facing a conflict between Meta’s contractual requirements and the rights of original creators.
The decision was made on an urgent basis. The judge said the harm could be irreversible once content had been used to train AI systems, and rejected the argument that publishers could protect themselves through the robots.txt file, which signals to web crawlers which pages should not be accessed.
Meta has not yet been heard in the case and is expected to respond in the coming weeks. The company said it would not comment. ANJ chief executive Marcelo Rech welcomed the injunction, calling it an initial step toward preventing platforms and AI models from appropriating third-party content.
Sources: Folha Mercado.