Brazilian Court Bars Meta From Using Local News Content to Train AI

Judge Guilherme Madeira Dezem of the 44th Civil Court granted the order on Tuesday, 3 November, according to The Rio Times and Moneynews. Three national press associations brought the case: ANJ, which represents newspapers, Abert, which represents radio and television broadcasters, and Aner, which represents magazine publishers. The associations argued that Meta's updated commercial terms made access to its platforms conditional on broad permission to use their journalism, including for AI.

What the Order Requires

According to published reports of the order, Meta may not use bots to collect articles, photos and videos from the websites of member outlets, and may not use that material to train or improve its AI systems. The order also bars Meta from penalising outlets that reject its new terms and requires it to keep their accounts, pages and tools active. The judge suspended the relevant part of the terms, which had been due to take effect on 20 November.

Non-compliance carries a fine of R$1 million a day, about US$182,000, capped at R$100 million, according to The Rio Times. Moneynews reported that the judge found "signs of possible copyright violations and contract overreach by Meta." The order applies only to members of the three associations, including regional outlets, and not to every news organisation in Brazil.

In a joint statement, the associations called the decision "an initial victory for the defence of intellectual property." They said it was the first known court ruling of its kind. Meta had not published a response or said whether it would appeal as of Wednesday morning, The Rio Times reported.

A Second Setback in a Week

The Brazilian order follows a ruling in Norway. On 28 October an Oslo court temporarily blocked the same terms after ten Norwegian media groups sued, according to The Rio Times. The terms apply to anyone who uses Meta's ads, pages and business tools, and continued use counts as acceptance. One clause lets Meta's automated crawlers access, index and analyse websites linked to a business account.

The Rio Times reported that Meta removed the word "train" from the wording and described the purpose as developing advertising and business products. Publishers argue that "develop" is broad enough to cover training AI models. Meta originally set a start date of 30 October and later moved it to 20 November after complaints from publishers and advertisers. A Meta spokesperson said the date could slip again if consultations with industry groups continue, and that site owners can still use robots.txt to keep crawlers off particular pages.

Why It Matters

Semafor reported that the decision adds to Brazil's record as one of Big Tech's most persistent adversaries. It came after President Luiz Inácio Lula da Silva accused Meta of favouring his challenger in the recent election. In 2024, Brazilian judges ordered a nationwide ban on X after Elon Musk rejected a Supreme Court order, and the national data protection authority stopped Meta from training generative AI on Brazilians' personal data.

For AI developers, the case is another test of whether platforms can obtain training rights through standard terms of service rather than through licences negotiated with publishers. Two national courts have now moved against Meta's approach within a week, and publishers in other markets are likely to watch both cases closely.