What happened
On 3 November 2026 a São Paulo civil court judge (44th Civil Court, case 4214874-21.2026.8.26.0100) granted a precautionary injunction, without hearing Meta, on a joint application from Brazil's ABERT, ANJ and ANER media associations. The judge suspended the clause in Meta's revised Commercial Terms (effective 20/11/2026) that would let Meta crawl business users' websites to build/train Meta AI, and ordered Meta to refrain from using any automated scraping tools or test credentials to access the members' domains or content, and from using that content to develop, train or feed Meta AI pending final judgment. It also ordered Meta to keep the members' Facebook, Instagram, WhatsApp and Threads accounts fully active without retaliation, under a daily penalty of BRL 1,000,000 capped at BRL 100,000,000. This is the second court order against the terms in a week, after an Oslo court blocked them on 28 October.
Why it matters
This is a binding judicial restriction on AI training-data collection and on platform terms that force users to accept scraping or give up access. It materially constrains how Meta (and, by precedent, other platforms) may collect web content for model training, and reinforces rising legal pushback — across Brazil, Norway and the EU — against implied-consent/accept-or-leave AI data harvesting.
Action needed
AI/platform companies should re-examine consent mechanisms for AI training-data collection of business users' and publishers' content — avoid by-default scraping and punitive 'accept-or-leave' terms — and review opt-out/contractual arrangements in Brazil and other jurisdictions facing similar injunctions.