When a regulator has to remove a feature before a company can prove that children can use it safely, the company’s safety system has already failed.
Brazil did not ban Discord. On August 12, the National Data Protection Authority ordered the company to suspend Go Live and equivalent video and screen sharing features for users in Brazil. The suspension remains in place until Discord demonstrates effective safeguards and receives permission to restore them. Discord confirmed Monday that video calls and screen sharing are unavailable, while direct messages, servers, voice channels and audio calls still work.
The order followed an investigation into an online group accused of coercing minors into violence and self-harm. A 13-year-old girl died in July after what authorities describe as a campaign that used Discord and other platforms. The public record does not establish that her death was broadcast on Discord, and Discord disputes that account. The dispute matters, but it leaves Discord with the same responsibility. The company offered private live video to minors without proving that its other safety controls could catch abuse before somebody was hurt.
Encryption is not an absolution
Discord has a real technical defense. Since March 2, the company has required end-to-end encryption for audio and video in direct messages, group calls, voice channels and Go Live streams. Only intended participants can decrypt those communications. Discord cannot simply inspect every private video call on its servers without breaking the privacy promise it made to users. That technical constraint should be protected because private conversations should stay private.
Brazil’s regulator is not entitled to turn child safety into a warrant for universal surveillance, and Brazilian law says so. The country’s Digital Statute for Children and Adolescents forbids mass, generic or indiscriminate surveillance. The same law requires services likely to be used by minors to manage foreseeable risks, provide safer defaults and prevent contact with material that encourages self-harm or suicide.
Discord wants the privacy half of that bargain without demonstrating the safety half. In its published findings, the authority says Discord depends on flawed automation and reports from people inside the same private groups where abuse occurs. The regulator also says the company’s measures mostly respond after harm or offer limited prevention. Those are preliminary administrative findings, not a court judgment. They are still a damning account of a platform asking the public to trust safeguards it could not show were working.
Encryption changes how safety must be built while leaving the duty intact.
A report after the harm is an incident log
Discord’s weakest defense is its dependence on users inside a server to report abuse. In an ordinary community, reporting can help. In a private group organized around coercion, the people with the report button may include the perpetrators. A frightened child should not be the last reliable sensor in a safety system built by a company serving more than 90 million people every day.
The warning record is long. SaferNet Brazil says it processed 406 Discord URLs in the first seven months of 2026, up from 264 during the same period last year. That is a 54 percent increase. Complaints are not convictions, and the number does not measure how common abuse is across Discord. It does show a worsening stream of warnings that the company and Brazilian authorities had reason to take seriously.
Police records add weight to that pattern. In February, investigators in Espírito Santo said they had apprehended a teenager suspected of participating in a digital criminal organization on Discord that allegedly circulated child sexual abuse material and encouraged self-harm. A 2025 national operation described groups using Discord, Telegram and other networks to attract young people and push them toward self-harm and extremist violence. These are allegations from active or prior investigations, not findings of guilt. They also make it impossible to pretend that closed abusive groups appeared from nowhere last month.
Brazil’s order is blunt because the failure is real
The strongest criticism of Brazil’s response is also correct: disabling video and screen sharing across an entire country is a blunt intervention. Discord says the campaign surrounding the girl’s death crossed multiple platforms and that it shut the relevant server before her death. Bad actors can move to another service or try to evade a geographic restriction. AP quoted network engineer Thiago Ayub warning that migration could push abusive users onto platforms receiving less scrutiny. Ordinary people lose a useful feature even though they did nothing wrong.
The government’s compressed timeline deserves scrutiny too. First lady Rosangela Lula da Silva called for Discord to be taken offline on August 6. The authority opened its inspection the next day, met with Discord on August 11 and issued the measure on August 12. Emergency action cannot become a substitute for careful technical regulation.
But the order is narrower than the proposed shutdown, reversible and subject to appeal. The measure targets live video and equivalent sharing tools, not private text or audio across the service. The authority also gave Discord a restoration test: demonstrate safeguards that work. That is a harder standard than publishing another statement about commitment to safety, which is precisely why the company objects to it.
Discord made this a product choice
Encryption can coexist with child safeguards without turning every user into an identity record. Discord already uses on-device age estimation and a third-party identity check in Brazil. The company says it receives only an age group and that people who decline verification can keep using ordinary messages, servers and voice chat. Discord’s own design accepts the principle that stronger controls can be attached to higher-risk features without building a government database of every conversation.
Discord could restrict live video for minors by default. Safety signals could run on the device or use account behavior instead of feeding private calls into a central scanner. A high-risk stream could be paused and routed to trained human reviewers. Servers with repeated child-safety reports could lose video privileges while an appeal is heard. Each approach would need careful testing, but waiting for someone inside an abusive group to file a report after the damage begins is worse.
Parents and police have responsibilities, and regulators can overreach. Discord still owns the product decisions that put private live video, minors and a report-driven safety model in the same room.
Privacy protects children too. Keep the encryption. Then build a safety system worthy of it, one that starts before the incident report.