This is the third post of a series about internet intermediary liability in Brazil. Our first post gives an overview of Brazil's current internet intermediary liability regime, set out in a law known as "Marco Civil da Internet," the context of its approval in 2014, and the beginning of the Supreme Court's judgment of such regime in November 2024. Our second post provides a bigger picture of the Brazilian context underlying the court's analysis and its most likely final decision.
The court’s examination of Marco Civil’s Article 19 began with Justice Dias Toffoli in November last year. We explained here about the cases under trial, the reach of the Supreme Court’s decision, and Article 19’s background related to Marco Civil’s approval in 2014. We also highlighted some aspects and risks of Justice Dias Toffoli’s vote, who considered the intermediary liability regime established in Article 19 unconstitutional.
Most of the justices have agreed to find this regime at least partially unconstitutional, but differ on the specifics. Relevant elements of their votes include:
Notice-and-takedown is likely to become the general rule for platforms' liability for third-party content (based on Article 21 of Marco Civil). Justices still have to settle whether this applies to internet applications in general or if some distinctions are relevant, for example, applying only to those that curate or recommend content. Another open question refers to the type of content subject to liability under this rule: votes pointed to unlawful content/acts, manifestly criminal or clearly unlawful content, or opted to focus on crimes. Some justices didn’t explicitly qualify the nature of the restricted content under this rule.
If partially valid, the need for a previous judicial order to hold intermediaries liable for user posts (Article 19 of Marco Civil) remains in force for certain types of content (or certain types of internet applications). For some justices, Article 19 should be the liability regime in the case of crimes against honor, such as defamation. Justice Luís Roberto Barroso also considered this rule should apply for any unlawful acts under civil law. Justice Cristiano Zanin has a different approach. For him, Article 19 should prevail for internet applications that don’t curate, recommend or boost content (what he called “neutral” applications) or when there’s reasonable doubt about whether the content is unlawful.
Platforms are considered liable for ads and boosted content that they deliver to users. This was the position held by most of the votes so far. Justices did so either by presuming platforms’ knowledge of the paid content they distribute, holding them strictly liable for paid posts, or by considering the delivery of paid content as platforms’ own act (rather than “third-party” conduct). Justice Dias Toffoli went further, including also non-paid recommended content. Some justices extended this regime to content posted by inauthentic or fake accounts, or when the non-identification of accounts hinders holding the content authors liable for their posts.
Monitoring duty of specific types of harmful and/or criminal content. Most concerning is that different votes establish some kind of active monitoring and likely automated restriction duty for a list of contents, subject to internet applications' liability. Justices have either recognized a “monitoring duty” or considered platforms liable for these types of content regardless of a previous notification. Justices Luís Roberto Barroso, Cristiano Zanin, and Flávio Dino adopt a less problematic systemic flaw approach, by which applications’ liability would not derive from each piece of content individually, but from an analysis of whether platforms employ the proper means to tackle these types of content. The list of contents also varies. In most of the cases they are restricted to criminal offenses, such as crimes against the democratic state, racism, and crimes against children and adolescents; yet they may also include vaguer terms, like “any violence against women,” as in Justice Dias Toffoli’s vote.
Complementary or procedural duties. Justices have also voted to establish complementary or procedural duties. These include providing a notification system that is easily accessible to users, a due process mechanism where users can appeal against content restrictions, and the release of periodic transparency reports. Justice Alexandre de Moraes also specifically mentioned algorithmic transparency measures.
Oversight. Justices also discussed which entity or oversight model should be used to monitor compliance while Congress doesn’t approve a specific regulation. They raised different possibilities, including the National Council of Justice, the General Attorney’s Office, the National Data Protection Authority, a self-regulatory body, or a multistakeholder entity with government, companies, and civil society participation.
Three other justices have yet to present their votes to complete the judgment. As we pointed out, the ruling will both decide the individual cases that entered the Supreme Court through appeals and the “general repercussion” issues underlying these individual cases. For addressing such general repercussion issues, the Supreme Court approves a thesis that orients lower court decisions in similar cases. The final thesis will reflect the majority of the court's agreements around the topics we outlined above.
Justice Alexandre de Moraes argued that the final thesis should equate the liability regime of social media and private messaging applications to the one applied to traditional media outlets. This disregards important differences between both: even if social media platforms curate content, it involves a massive volume of third-party posts, mainly organized through algorithms. Although such curation reflects business choices, it does not equate to media outlets that directly create or individually purchase specific content from approved independent producers. This is even more complicated with messaging applications, seriously endangering privacy and end-to-end encryption.
Justice André Mendonça was the only one so far to preserve the full application of Article 19. His proposed thesis highlighted the necessity of safeguarding privacy, data protection, and the secrecy of communications in messaging applications, among other aspects. It also indicated that judicial takedown orders must provide specific reasoning and be made available to platforms, even if issued within a sealed proceeding. The platform must also have the ability to appeal the takedown order. These are all important points the final ruling should endorse.
Risks and Blind SpotsWe have stressed the many problems entangled with broad notice-and-takedown mandates and expanded content monitoring obligations. Extensively relying on AI-based content moderation and tying it to intermediary liability for user content will likely exacerbate the detrimental effects of these systems’ limitations and flaws. The perils and concerns that grounded Article 19's approval remain valid and should have led to a position of the court preserving its regime.
However, given the judgement’s current stage, there are still some minimum safeguards that justices should consider or reinforce to reduce harm.
It’s crucial to put in place guardrails against the abuse and weaponization of notification mechanisms. At a minimum, platforms shouldn’t be liable following an extrajudicial notification when there’s reasonable doubt concerning the content’s lawfulness. In addition, notification procedures should make sure that notices are sufficiently precise and properly substantiated indicating the content’s specific location (e.g. URL) and why the notifier considers it to be illegal. Internet applications must also provide reasoned justification and adequate appeal mechanisms for those who face content restrictions.
On the other hand, holding intermediaries liable for individual pieces of user content regardless of notification, by massively relying on AI-based content flagging, is a recipe for over censorship. Adopting a systemic flaw approach could minimally mitigate this problem. Moreover, justices should clearly set apart private messaging applications, as mandated content-based restrictions would erode secure and end-to-end encrypted implementations.
Finally, we should note that justices generally didn’t distinguish large internet applications from other providers when detailing liability regimes and duties in their votes. This is one major blind spot, as it could significantly impact the feasibility of alternate and decentralized alternatives to Big Tech’s business models, entrenching platform concentration. Similarly, despite criticism of platforms’ business interests in monetizing and capturing user attention, court debates mainly failed to address the pervasive surveillance infrastructure lying underneath Big Tech’s power and abuses.
Indeed, while justices have called out Big Tech’ enormous power over the online flow of information – over what’s heard and seen, and by whom – the consequences of this decision can actually deepen this powerful position.
It’s worth recalling a line of Aaron Schwarz in the film “The Internet’s Own Boy” when comparing broadcasting and the internet. He said: “[…] what you see now is not a question of who gets access to the airwaves, it’s a question of who gets control over the ways you find people.” As he puts it, today’s challenge is less about who gets to speak, but rather about who gets to be heard.
There’s an undeniable source of power in operating the inner rules and structures by which the information flows within a platform with global reach and millions of users. The crucial interventions must aim at this source of power, putting a stop to behavioral surveillance ads, breaking Big Tech’s gatekeeper dominance, and redistributing the information flow.
That’s not to say that we shouldn’t care about how each platform organizes its online environment. We should, and we do. The EU Digital Services Act, for example, established rules in this sense, leaving the traditional liability regime largely intact. Rather than leveraging platforms as users’ speech watchdogs by potentially holding intermediaries liable for each piece of user content, platform accountability efforts should broadly look at platforms’ processes and business choices. Otherwise, we will end up focusing on monitoring users instead of targeting platforms’ abuses.
This is the second post of a series about intermediary liability in Brazil. Our first post gives an overview of Brazil's current intermediary liability regime, the context of its approval in 2014, and the beginning of the Supreme Court's analysis of such regime in November 2024. Our third post provides an outlook on justices' votes up until June 23, underscoring risks, mitigation measures, and blind spots of their potential decision.
The Brazilian Supreme Court has formed a majority to overturn the country’s current online intermediary liability regime. With eight out of eleven justices having presented their opinions, the court has reached enough votes to mostly remove the need for a previous judicial order demanding content takedown to hold digital platforms liable for user posts, which is currently the general rule.
The judgment relates to Article 19 of Brazil’s Civil Rights Framework for the Internet (“Marco Civil da Internet,” Law n. 12.965/2014), wherein internet applications can only be held liable for third-party content if they fail to comply with a judicial decision ordering its removal. Article 19 aligns with the Manila Principles and reflects the important understanding that holding platforms liable for user content without a judicial analysis creates strong incentives for enforcement overreach and over censorship of protected speech.
Nonetheless, while Justice André Mendonça voted to preserve Article 19’s application, four other justices stated it should prevail only in specific cases, mainly for crimes against honor (such as defamation). The remaining three justices considered that Article 19 offers insufficient protection to constitutional guarantees, such as the integral protection of children and teenagers.
The judgment will resume on June 25th, with the three final justices completing the analysis by the plenary of the court. Whereas Article 19’s partial unconstitutionality (or its interpretation “in accordance with” the Constitution) seems to be the position the majority of the court will take, the details of each vote vary, indicating important agreements still to sew up and critical tweaks to make.
As we previously noted, the outcome of this ruling can seriously undermine free expression and privacy safeguards if they lead to general content monitoring obligations or broadly expand notice-and-takedown mandates. This trend could negatively shape developments globally in other courts, parliaments, or with respect to executive powers. Sadly, the votes so far have aggravated these concerns.
But before we get to them, let's look at some circumstances underlying the Supreme Court's analysis.
2014 vs. 2025: The Brazilian Techlash After Marco Civil's ApprovalHow did Article 19 end up (mostly) overturned a decade after Marco Civil’s much-celebrated approval in Brazil back in 2014?
In addition to the broader techlash following the impacts of an increasing concentration of power in the digital realm, developments in Brazil have leveraged a harsher approach towards internet intermediaries. Marco Civil became a scapegoat, especially Article 19, within regulatory approaches that largely diminished the importance of the free expression concerns that informed its approval. Rather than viewing the provision as a milestone to be complemented with new legislation, this context has reinforced the view that Article 19 should be left behind.
The tougher approach to internet intermediaries gained steam after former President Jair Bolsonaro’s election in 2018 and throughout the legislative debates around draft bill 2630, also known as the “Fake News bill.”
Specifically, though not exhaustive, concerns around the spread of disinformation, online-fueled discrimination, and political violence, as well as threats to election integrity, constitute an important piece of this scenario. This includes the use of social media by the far right within the escalation of acts seeking to undermine the integrity of elections and ultimately overthrow the legitimately elected President Luis Inácio da Silva in January 2023. Investigations later unveiled that related plans included killing the new president, the vice-president, and Justice Alexandre de Moraes.
Concerns over child and adolescents’ rights and safety are another part of the underlying context. Among others, a wave of violent threats and actual attacks in schools in early 2023 was bolstered by online content. Social media challenges also led to injuries and deaths of young people.
Finally, the political reactions to Big Tech’s alignment with far-right politicians and feuds with Brazilian authorities complete this puzzle. It includes reactions to Meta’s policy changes in January 2025 and the Trump’s administration’s decision to restrict visas to foreign officials based on grounds of limiting free speech online. This decision is viewed as an offensive against Brazil's Supreme Court from U.S. authorities in alliance with Bolsonaro’s supporters, including his son now living in the U.S.
Changes in the tech landscape, including concerns about the attention-driven information flow, alongside geopolitical tensions, landed in Article 19 examination by the Brazilian Supreme Court. Hurdles in the legislative debate of draft bill 2630 turned attention to the internet intermediary liability cases pending in the Supreme Court as the main vehicles for providing “some” response. Yet, the scope of such cases (explained here) determined the most likely outcome. As they focus on assessing platform liability for user content and whether it involves a duty to monitor, these issues became the main vectors for analysis and potential change. Alternative approaches, such as improving transparency, ensuring due process, and fostering platform accountability through different measures, like risk assessments, were mainly sidelined.
Read our third post in this series to learn more about the analysis of the Supreme Court so far and its risks and blind spots.
Like users of all technologies, ChatGPT users deserve the right to delete their personal data. Nineteen U.S. States, the European Union, and a host of other countries already protect users’ right to delete. For years, OpenAI gave users the option to delete their conversations with ChatGPT, rather than let their personal queries linger on corporate servers. Now, they can’t. A badly misguided court order in a copyright lawsuit requires OpenAI to store all consumer ChatGPT conversations indefinitely—even if a user tries to delete them. This sweeping order far outstrips the needs of the case and sets a dangerous precedent by disregarding millions of users’ privacy rights.
The privacy harms here are significant. ChatGPT’s 300+ million users submit over 1 billion messages to its chatbots per day, often for personal purposes. Virtually any personal use of a chatbot—anything from planning family vacations and daily habits to creating social media posts and fantasy worlds for Dungeons and Dragons games—reveal personal details that, in aggregate, create a comprehensive portrait of a person’s entire life. Other uses risk revealing people’s most sensitive information. For example, tens of millions of Americans use ChatGPT to obtain medical and financial information. Notwithstanding other risks of these uses, people still deserve privacy rights like the right to delete their data. Eliminating protections for user-deleted data risks chilling beneficial uses by individuals who want to protect their privacy.
This isn’t a new concept. Putting users in control of their data is a fundamental piece of privacy protection. Nineteen states, the European Union, and numerous other countries already protect the right to delete under their privacy laws. These rules exist for good reasons: retained data can be sold or given away, breached by hackers, disclosed to law enforcement, or even used to manipulate a user’s choices through online behavioral advertising.
While appropriately tailored orders to preserve evidence are common in litigation, that’s not what happened here. The court disregarded the privacy rights of millions of ChatGPT users without any reasonable basis to believe it would yield evidence. The court granted the order based on unsupported assertions that users who delete their data are probably copyright infringers looking to “cover their tracks.” This is simply false, and it sets a dangerous precedent for cases against generative AI developers and other companies that have vast stores of user information. Unless courts limit orders to information that is actually relevant and useful, they will needlessly violate the privacy rights of millions of users.
OpenAI is challenging this order. EFF urges the court to lift the order and correct its mistakes.
