On 5 October 2026, Brazil's data protection authority (ANPD) opened a public page for complaints about internet application providers. Anyone can file, and anonymous complaints are accepted if they carry enough evidence. The channel covers missing notice channels for illegal content, weak safeguards against ads or boosting of such content, missed removal deadlines, and failures to protect against the spread or creation of intimate content. It is the enforcement route for Decrees 12.975/2026 and 12.976/2026. It is not limited to gender-based violence.
The case for the new regime
The strongest argument for the decrees is that notice-and-takedown without deadlines fails victims of non-consensual intimate imagery. The harm compounds by the hour, and AI tools now let anyone fabricate such images. Decree 12.976/2026, published 21 May 2026, responds directly. Article 7 requires unauthorised intimate content to be made unavailable within two hours of a notice. It also requires removed material to be flagged to block re-uploads and a permanent, free reporting channel. Article 3 defines intimate content to include material created or altered by AI, and Article 10 requires AI services to build safeguards against prohibited requests, scaled to their size and risk.
The complaints channel also followed a public failure. Press coverage says ANPD was criticised for allegedly acting too slowly against Discord after a 13-year-old girl died by suicide during a livestream. A judge called the agency's conduct an "omissão". ANPD replied that it suspended Discord's broadcasts five days after opening an enforcement proceeding. A regulator under that pressure needs an intake route, and a public one is better than ad hoc escalation.
What the architecture gets right
Several design choices are proportionate. Decree 12.975/2026 does not make platforms liable for a single missed item. Article 16-B ties liability to a "falha sistêmica", which arises when a provider cannot show adequate prevention or removal measures. Article 16-B §3 says isolated unlawful content is not by itself a systemic failure, and Article 16-I bars liability based solely on keeping or removing one notified item. Article 16-D voids notices that do not identify the conduct, the specific content and the notifier. That blunts mass-flagging campaigns. Article 16-E requires platforms to tell both the notifier and the uploader about a decision, with reasons and a way to contest.
The channel itself is also restrained. ANPD says it cannot moderate content or order removal of specific posts, it does not send individual replies, and it does not investigate crimes. Complainants should first go to the platform, and criminal matters go to the police. That limits the risk of a regulator becoming a central censor.
Where the risks sit
The first risk is the clock. Decree 12.976 sets transitional deadlines in Article 12: six hours for manifestly illegal content and 24 hours otherwise, either to remove or to explain why the content stays up. Contested decisions must be resolved within 24 hours. Two hours for intimate imagery is defensible, because the unlawfulness is usually obvious and the victim is identifiable. The risk is that fixed clocks applied across wider categories push platforms to remove first and review later. The decree does allow providers to keep content they reasonably doubt is criminal, if they explain why. That safe harbour only works if ANPD treats a documented refusal as compliance rather than as a failure.
The second risk is the vagueness of the systemic-failure test. Because Decree 12.975 leaves the removal period for later regulation, with Article 16-D delegating it to the competent authority, a provider cannot know precisely what "adequate" means. Large platforms can afford compliance teams and automated filters. Smaller services and federated networks may respond by over-blocking or by exiting the market. EFF's October comments to ANPD make the same point. They call for solid appeal routes for restricted posts, less reliance on AI moderation, narrow application of duty-of-care and systemic-failure concepts, and no normalisation of blocking platforms.
The third risk is the volume and quality of complaints. ANPD says it monitors 22 platforms, app stores and generative AI tools. An open channel that takes anonymous reports without individual replies could generate a large backlog. The agency would then triage by public pressure rather than by risk. Political attention could pull enforcement toward high-profile cases while routine failures to answer victims go unaddressed.
What ANPD should do next
Three steps would keep enforcement proportionate and effective:
- Publish enforcement criteria. ANPD should state what evidence of "adequate measures" will satisfy Article 16-B, including for smaller providers.
- Report on the channel. Aggregate data on complaints received, how many led to a monitoring proceeding, and outcomes would show whether the channel targets real harm.
- Protect appeals. Platform decisions that reject or remove content should carry a working contest route, as Article 16-E already requires, and ANPD should audit it.
Brazil has chosen a regulator-led model for platform duties. Done well, it can protect women from some of the fastest-moving forms of abuse, including AI-made intimate images. Done poorly, it will reward over-removal. The complaints channel is a sensible start, but its value depends on how ANPD defines adequate measures and whether it measures its own results.