The Cost of Moderation: STF’s Impact on Digital Platform Liability and the Justice System

Reglab is a think tank specializing in research and consultancy that assists companies, business associations and policymakers with data-driven planning and impact analysis. Our focus is on making responsible and strategic decisions, unraveling the regulatory challenges of the media and technology sector.

Our goal is to promote evidence-based research that increases accountability and establishes meaningful milestones and goals for the ecosystem.

Find out more at www.reglab.com.br

The Policy Briefs Series encompasses studies that evaluate existing or proposed public policies, using qualitative and quantitative data to inform and guide strategic decisions. The objective is to bring complex issues into an accessible way, highlighting the main points of analysis, impacts and possible recommendations.

Executive director: Pedro Henrique Ramos

Authors: Daniela Naomi Shimabukuro Nomura, João Ricardo Costa Filho and Pedro Henrique Ramos

Researcher: Victória Dean Figueiredo Ribeiro

Final Layout: Eliza Natsuko Shiroma

Suggested citation: NOMURA, D. N. S.; COSTA FILHO, J. R.; RAMOS, P. H. The Price of Moderation: Impacts on the Judiciary and the Debate on the Revision of the Marco Civil da Internet by the STF. Policy Briefs Reglab, n. 2. São Paulo: Reglab, 2025.

Executive Summary

A historic decision about the internet can generate unexpected effects for the

justice system. The Federal Supreme Court (STF) is close to resuming the judgment of Issues 987 and 533, which assess the constitutionality of article 19 of the Marco Civil da Internet (MCI). The votes already presented by ministers Toffoli, Fux and Barroso indicate different paths, with potential impacts not only on

content moderation, but also on the structural dynamics of the internet in Brazil — with side effects still

poorly sized for the digital ecosystem and the functioning of the public sector.

Faced with the possibility of adopting a new objective liability regime for digital platforms, this study analyzes an externality that is rarely addressed in the debate: What are the budgetary impacts of a change of this type for the Judiciary itself?

To do this, we use econometric models based on synthetic control and scenario projections for the period from 2025 to 2029, based on real data on judicialization and procedural costs. From this, we projected four scenarios for the period of 2025 to 2029, varying according to the degree of responsibility and the intensity of the propagation of effects.

Main findings:

754 thousand

Number of new lawsuits over the next 5 years if there is a shift to strict liability

160%

This is the increase in shares in relation to the current scenario (2.6x), with effects

potentially concentrated in civil courts and special courts

R$777 million

This is the most serious scenario with a financial impact on the Judiciary, an amount equivalent to hiring 1,600 federal judges for one year

Actions involving content moderation may compromise the efficiency of the Judiciary, as they are, in general, more complex than typical consumer disputes and require more time and effort to judge, which can make these values higher;

The risk of abusive litigation is high, especially if the STF decision

is broad, poorly modulated and without safeguards, fostering a predatory “stock industry” and causing losses to the Judiciary’s accounts;

The adoption of alternative regimes based on subjective responsibility – with specific delimitation of hypotheses of low discretion

and systemic obligations –, following international trends, can considerably reduce the financial impacts on the Judiciary.

Introduction

In November 2024, the STF Plenary began judging Topics 987 and 533, arising from Extraordinary Appeals (RE) Nos. 1,037,396 and 1,057,258 reported by ministers Dias Toffoli and Luiz Fux. The cases discuss the civil liability of digital platforms for damage caused by third-party content and, consequently, the constitutionality of article 19 of the MCI.

So far, three of the eleven STF ministers have voted. Rapporteurs Dias Toffoli and Luiz Fux and President Luís Roberto Barroso converge to declare the unconstitutionality of article 19 of the MCI and propose new theses to modify the accountability of platforms. However, the three have very different proposals, which makes the search for consensus challenging.

The trial is suspended following a request from Minister André Mendonça on 12/18/2024. This decision could change the understanding of the responsibility of digital platforms, with impacts on freedom of expression and the diversity of content on the internet.

It is a debate with many perspectives: in 2024, Reglab published the study “Interesses em Jogo no STF: Mapping the Dispute for the Marco Civil da Internet”, which demonstrated

the complexity of the discussion: more than 56 entities spoke out in the process, presenting 63 unique arguments (Ramos et al., 2024).

A little explored aspect is the economic consequences of this decision for the Public Sector: depending on the thesis that prevails in the STF, how will this impact the Judiciary and public accounts? From this specific section, we intend to present new elements to enrich the debate.

How does the responsibility of digital platforms work?

Digital platforms – here understood as services that facilitate or mediate interactions between two or more agents via the internet (CADE, 2023) – create spaces for their users to create and share content (Teffé, 2024). With the high number of users of some of them, such as social media services, questions arise about the platform’s degree of responsibility for the content of these third parties.

In Brazil, this issue is currently addressed in Law 12,965/14, known as the Civil Rights Framework for the Internet. Currently, the current rule is subjective responsibility, but with specific criteria that go beyond guilt or intent – what we call judicial notice and take down. It works like this:

• Internet platforms are not automatically responsible for the content published by users.
• Platforms can only be held responsible after receiving a specific court order to remove content and failing to comply with it.
• This does not prevent platforms, based on their own moderation policies, from being able to remove content in advance, either by their own decision or by notification from the user.

A CASE OF APPLICATION OF THE MCI MODELIn 2022, the Superior Court of Justice (STJ) evaluated a case in which a person published criticism of the appointment of another to the Public Security Secretariat of the State of São Paulo on their social network. The publication mentioned that he was involved in a corruption scheme. The platform did not remove the publication and, subsequently, the public agent requested the content to be removed from the Judiciary, alleging defamation. The STJ decided that the post did not constitute an illegal act because it was political criticism based on credible facts and of public interest, even if expressed in a severe, ironic or merciless manner (STJ, 2022). This case exemplifies the importance of article 19 in promoting public debates, and the importance of the Judiciary in balancing fundamental rights.

This rule, when established in 2014, aimed to strike a balance between freedom of expression and the protection of individual rights. But there are two exceptions: for content that violates copyright or involves non-consensual intimate photos, platforms must remove the material as soon as they are notified directly by the victim, dispensing with a court order – this is what we call notice and take down: responsibility remains subjective, but does not require a court decision, simply failing to comply with the user’s request is enough for the platform to be held responsible.

What is being discussed in the STF Judgment?

With the start of the judgment of Topics 987 and 533 by the STF, the ministers proposed new theses and rules for the platforms’ responsibility. Barroso, Toffoli and Fux support the need to change the judicial notice and takedown fornotice and takedown, but with different approaches:

Barroso defends maintaining the subjective liability regime, but limiting the procedure for judicial notice and take down to allegations of crimes against honor, civil offenses and residual content – for the rest, the notice and take down would be applicable.

Toffoli and Fux proposed broad theses that, in practice, change the procedure and alter it to objective liability, that is, regardless of notification1. Under this regime, the platforms’ liability would be automatic, regardless of fault, based on the obligation to compensate for the risk of the activity (Miragem, 2021).

• In his vote, Minister Luiz Fux proposed the thesis of absolute presumption of knowledge of the illegality of content, that is, in which there is no need to prove the platform’s guilt. According to the vote, in cases of posts promoted for payment, it is presumed that

It is clear that the company is aware of the content, making prior notification for civil liability unnecessary. In this way, even if we talk about “guilt” (in the sense of the science of illegality), it is presumed absolutely, approaching, in practical effects for the methodological purposes of this work, objective responsibility.

Strict liability is not new in Brazil: it occurs by legal provision and in cases such as consumer rights and environmental liability. However, its application to topics related to freedom of expression rights would be unprecedented.

Subjective liability through judicial examination (Judicial Notice and Takedown)

Subjective liability after knowledge (Notice and Takedown)

When is there accountability?

After failure to comply with court order

After direct request from the user and/or authority, without the need for a court decision

Regardless of notification or court decision

Who decides removal?

JudiciaryPlatformPlatform

How other countries address this issue

The debate about responsibility in content moderation is not exclusive to Brazil

— not even recent. It started in the United States in 1996, with the Section 230 of Communications Decency Act, which granted broad immunity to digital platforms for content generated by third parties, protecting them both through user publications and through their moderation decisions (Teffé; Souza, 2024).

Since the 2010s, this regime has been criticized, with reform proposals to balance the protection of platforms with more effective accountability mechanisms. So much so that the MCI, approved in 2014 in Brazil, started from a different logic:

rejected early immunity and opted for procedural rules of liability, making the removal of content conditional on the existence of a court order — the model known as judicial notice and takedown.

In 2020, European countries began to review regulatory frameworks, some of them questioning some of the premises set out in the Marco Civil. O Digital Services Act (DSA), in force since 2024 in European Union, maintained the model of subjective responsibility based on specific knowledge and strengthened safeguards to freedom

expression, prohibiting general monitoring obligations. While the DSA does not directly address civil liability in Member States, it establishes minimum principles that limit platform accountability while requiring systemic and transparency obligations.

No United Kingdom, the Online Safety Act (OSA), approved in 2023, also adopts subjective responsibility as a rule, but imposes proactive obligations to deal with illegal and harmful content, especially for the protection of children, requiring a higher level of regulatory diligence than the Marco Civil, with a focus on effective prevention and response systems — and not on accountability for individual content.

That is, the The evolution of regulation on content moderation in comparative law does not seem to change the general liability regime, in which the main paradigm continues to be subjective liability. The adoption of strict liability for digital platforms is not common – on the contrary, it is very rare and exceptional (Mello, 2024). China and Iran, for example, impose objective liability regimes for digital platforms due to content that is harmful to the image of the party and government (Wang, 2022) and that can “generate crimes” (Article 19, 2012), respectively, being among the few global cases that adopt this model.

The adoption of strict liability involves a series of effects — such as the risk of excessive content removals, impacts on freedom of expression and the transformation of moderation dynamics. These are topics that require in-depth study and research to understand their scope and effects, and how justified they are for possible social benefits.

Along these lines, this study proposes a very specific and little explored approach: the direct consequences of this regulatory change for the Judiciary itself.

Impacts of regulatory choices on the Judiciary

Experiences in other sectors show how regulatory changes influence the burden of the Judiciary, with effects on public accounts and on the actual access to justice.

In the Labor Court, a 2021 STF decision declared unconstitutional a labor reform rule that determined the obligation to pay expert and legal fees to the defeated party, regardless of income (STF, 2021). As a result, an incentive was created for beneficiaries of Eratuita Justice to file lawsuits, causing an increase of 378 thousand legal demands (Eercina, 2025). In the health area, the Government has been facing for decades a significant volume of actions regarding payment for medicines not included in the SUS list, which recently motivated an inter-federative agreement seeking to financially balance the public health system and reduce the overload of the Judiciary with individual demands (Reis, 2024).

It is important to emphasize that the increase in lawsuits or costs for the Judiciary is not, in itself, something negative. In many cases, the growth in litigation may reflect greater

raising awareness of rights, expanding access to justice or correcting historical asymmetries. What What we seek to discuss is not the legitimacy of these effects, but rather the extent to which they are anticipated, measured and considered at the time of decision making.. This is what we want to explore in this article. It seems logical that the statement

of unconstitutionality of article 19 of the MCI and the adoption of objective liability for platforms may encourage a significant increase in legal disputes seeking compensation for damages (Murta Eoyanes, 2023). But how much does this increase represent? How can its budgetary effects be reduced or increased by the STF decision?

We reinforce, once again, that the importance of the normative discussion on the responsibility of platforms is not denied here, especially in the face of impacts not only on people’s lives, but also on democracy and freedom of expression (Cruz, 2024), nor do we aim to discourage the use of the Judiciary as a space for claiming rights.

By quantifying potential impacts on the Judiciary, what we are looking for here is to contribute so that decisions with general repercussions can be accompanied by more complete assessments and strategies and evidence-based, anticipating unwanted side effects and ensuring that the final approved model is applicable and balanced, also taking into account the systemic impacts on the Justice system.

Methodology Summary

The analysis was limited to estimate the impact on the public budget of the Judiciary resulting from a possible adoption of a new liability regime

of digital platforms for content published by third parties, notably objective liability (as voted by Minister Dias Toffoli) or presumption of absolute guilt (vote by Minister Fux)2.

The impact on the public budget can come through two channels: (i) change in the quantity of processes and (ii) costs of these processes. Even though greater complexity in the processes may mean analyzes take longer and in more instances, we prefer to keep the cost variable as constant, based on the methodological hypothesis that the STF’s decision in Topics 987 and 533 will only have an impact on quantity. Therefore, the empirical challenge is to assess whether there will be a change in the number of actions in the Judiciary as a result of this STF decision.

To approximate the causal effect, the identification strategy considered a counterfactual scenario based on the regime currently in force, based on the hypothesis that, if the implementation of the MCI had adopted an objective liability regime, the volume of demands would have been higher. To test this hypothesis, the synthetic control model was used3 (Abadie; Eardeazabal, 2003; Abadie; Diamond; Hainmueller, 2010; Abadie, 2021).

Using real data from legal proceedings, a counterfactual simulation was constructed comparing the behavior of a “synthetic company” (which we call Company S), formed based on companies that act with the end public and that are subject to objective liability rules (e.g. product suppliers subject to the Consumer Protection Code), but that were not affected by the introduction of the MCI. This made it possible to estimate, based on historical evidence, how many more processes could have occurred in a scenario without the limitations of liability created from article 1G.

Based on this estimated difference, we project the additional volume of cases that different platforms could face between 2025 and 2029, if the STF changes the current regime. Based on the average cost of processing a lawsuit, we also calculate the budgetary impact for the Judiciary in different scenarios. The complete methodology, with the databases and parameters used, is detailed in the

end of the study.

• As presented in note 1, we understand that, even though they are proposals based on different legal institutes, their practical effects are similar and, therefore, this thesis adopts both votes as equivalent for methodological purposes.
• Synthetic control is a consolidated technique for evaluating the effect of public and economic policies, as well as changes in the legal framework. The scenarios built from the model do not constitute projections, but conditional simulations to guide public decisions based on evidence.

Results

THE METHODOLOGY, SIMPLIFIEDImagine that the Marco Civil da Internet had never existed and that, since 2014, Brazil had adopted a different rule – similar to the one now being discussed by the Supreme Court. To understand how this could have changed things, we compared digital platforms with other companies that work with the end public and that follow stricter liability rules – such as consumer companies. This is called “synthetic control” — it’s like creating an alternative version of history, using real data and scientifically proven mathematical formulas. To make these calculations, we look at the history of lawsuits and costs of the Judiciary, and we also include other important variables — such as people’s greater access to the internet and the emergence of new social networks. With this simulation, we were able to estimate how many more cases could have appeared in court — and how much this could cost the public coffers.

• Statistical Effects

– With or Without the Marco Civil da Internet

The pre-treatment period was defined between 2011 and 2014 and the construction of the weights in the creation of the synthetic control group was based on the simplex method. In the analysis

counterfactual, if the MCI had not been approved, it is estimated that the synthetic control group, the Company S, there would be a greater number of lawsuits. The following figure summarizes the results.

On the curve in pink, The behavior of the synthetic control group before and after treatment (the MCI) is shown. It is noted that, in the pre-treatment period, the control group has a performance similar to the observed data regarding the number of processes in the company that constitutes the treatment group. (the blue curve). Considering only the point estimate, the increase in the number of processes, if the MCI had not been approved, would reach, in the third year, around 51%.

The above results are statistically significant from the second year after treatment. The confidence intervals were constructed based on Cattaneo, Feng and Titiunik (2021), assuming a confidence level of 95%, and are represented in the figure below.

With interval estimation, the upper limit reaches a 160% increase in the number of processes in relation to the scenario with treatment.

In other words, the estimated increase in the number of content removal processes, if an objective liability rule is adopted, is 2.6 times the current volume.

Scenario Construction

Based on these results, it is possible to create some scenarios for the impacts on public spending on the Judiciary based on the STF’s decision on topics 533 and 987. It is important to emphasize that the scenarios are not predictive in nature, but exploratory and comparative, so that it is possible to map possible trajectories based on empirical evidence and the counterfactual model.

To construct econometric scenarios, it is necessary to estimate the trajectory of the number of shares of the main companies (e.g. Meta/Facebook, Youtube, TikTok, X/Twitter) under the

current conditions, that is, without the STF decision, for the period from 2025 to 202G. An ARIMA model (whose order was selected based on the AIC information criterion) was estimated for each of the time series of justice processes for all companies in the sector.

In addition to the projections for the base scenario for the number of actions, two choices were made to construct the scenarios. The first concerns the size of the impact of the STF decision. To this end, two types of counterfactual scenarios were considered: average increase and increase in upper limit of the number of shares, based on estimates from the synthetic control model.

Another premise used was about the pace of growth in the number of processes, also with two possibilities:

• MCI rhythm: the rate of percentage growth in the number of cases is similar to that observed in the counterfactual exercise of the synthetic control model year after year – that is, observing effects from 2015, 2016 and so on4;
• Peak MCI: the other in which in all years the counterfactual increase will be equal to the highest annual value estimated by the synthetic control model. This is the so-called maximum stress hypothesis, or upper-bound estimate (Firpo and Possebom, 2018) and aims to test the impacts of a possible immediate litigious explosion, without initial containment mechanisms.

From these combinations, four scenarios then emerge:

Scenario 1 (average increase, MCI pace)

Progressive growth in the volume of shares, similar to the pace observed after the MCI came into force.

158.403165,13

Scenario 2 (average increase, peak MCI)

O volume anual de ações replica, de forma constante, o pico observado após o MCI, simulando uma explosão litigiosa já em 2025.

243.476250,88

Scenario 3

(upper threshold, MCI rhythm)

Supõe alta litigância, mas com crescimento distribuído ao longo dos anos.

584.447604,20

Scenario 4

(upper threshold, MCI peak)

Maior volume possível de ações desde o primeiro ano, sem medidas de contenção, replicando anualmente o pior cenário já registrado.

754.683777,62

• isto é, a quantidade de processos em relação às projeções com os modelos ARIMA seria 22,36% superior no primeiro ano após a mudança do entendimento do STF, 42,02% superior no segundo ano, 51,56% superior no terceiro ano, 37,35% superior no quarto ano e 17,99% superior no quinto ano, quando consideramos o estimador médio (ou 91,48%, 137,09%, 159,82%, 135,6% e 95,16%, respectivamente, quando o limiar superior for utilizado)

A Tabela acima apresenta os resultados das simulações em cada um dos quatro cenários no que tange ao aumento na quantidade de ações acumulado ao longo de cinco anos e o valor presente dos gastos extras por parte do Poder Judiciário que a maior quantidade de processos representaria.

Considerando apenas esses quatro cenários, estima-se um aumento de até 754 mil ações entre 2025 e 2029, seguido de custo para o Judiciário de até 777,62 milhões de reais.

Results Discussion

Os resultados são consistentes com a opinião já apresentada por especialistas jurídicos, que apontam que o aumento de processos judiciais decorre do fato de que, sob um regime de “responsabilidade automática” das plataformas, há uma dinâmica que tende a fomentar demandas judiciais por parte de indivíduos ofendidos por publicações, mas também por usuários que tenham seus conteúdos retirados do ar de forma indevida (Ramalho, 2024).

As teses propostas pelos ministros Dias Toffoli e Luiz Fux estabelecem uma série de situações em que a responsabilidade entraria nesse cenário e, logo, criando um incentivo para pedidos de indenização mesmo antes de uma denúncia direta à plataforma. No voto de Dias Toffoli, essas hipóteses incluiriam:

• Inauthentic profiles
• Copyright infringement
• Conteúdos moderados, recomendados ou impulsionados pelos provedores
• Crimes against the Democratic Rule of Law
• Acts of terrorism or preparatory to terrorism
• Crime of inducing, instigating or assisting suicide
• Crime of racism
• Any type of violence against children, adolescents and vulnerable people
• Any kind of violence against women
• Health infraction, for failing to execute, hindering or opposing the execution of health measures in a Public Health Emergency of National Importance
• Human trafficking
• Incitement or threat of acts of physical or sexual violence
• Disclosure of notoriously untrue or seriously out of context facts that

lead to incitement to physical violence, threats to life or acts of violence against groups or members of socially vulnerable groups; and

• Disclosure of facts that are notoriously untrue or out of context with the potential to cause damage to the balance of the claim or the

integrity of the electoral process

Although, according to the vote, these categories are exhaustive, it is important to highlight that are formulated in a sufficiently broad manner, which leaves a high margin for divergent interpretations.

This dynamic creates a favorable environment for the development of a “compensation industry” (Ramalho, 2024), in which judicialization becomes an instrument

economically attractive for abusive actions aimed at obtaining financial compensation, without due analysis of guilt. We also draw attention to the actions of service companies, similar to those operating in the civil aviation sector, which would encourage litigation by hundreds of users who have had their posts removed (Ramalho, 2024).

“litigation against technology companies can become an even more advantageous business than opening a technology company” (Souza, in Ramalho, 2024).

In this context, fraudulent cases of creating supposedly false profiles, intentional posting of offensive content and self-implication schemes, in which the author of the content pretends to be the victim, are examples of strategies that will be used by litigants in bad faith in an undue search for pecuniary compensation, which can contribute to an even greater overload of the Judiciary (Batista, 2024).

Overload, Delays and Access to Justice

The increase in litigation compromises the efficiency of the judicial services provided (CIJMG, 2022). As mentioned by Minister Luís Roberto Barroso, the loss in the quality of judicial provision often translates into an increase in the rate of errors, which leads to the production of contradictory decisions on the same subject and encouraging non-compliance with precedents (CIJMG, 2022). The lack of uniformity between precedents also contributes to the formation of unstable and unpredictable jurisprudence, which ends up stimulating the multiplication of legal demands (Eimenes, 2024).

35 millionNew cases in 2024, 22 million of which originate 4 years and 5 monthsAverage processing time for a case in State Court

Total number of ongoing legal cases in Brazil

Data from Justice in Numbers (National Council of Justice, 2024)

Therefore, even though the number of 750 thousand actions over a period of five years is not statistically relevant within the general numbers of the Brazilian Justice, its increase of 160% is exponentially higher than what is observed in the percentage variation between 2022 and 2023 in the Judiciary, which was 9.4% (CNJ, 2024).

Still, it is necessary to highlight the complexity of the actions: compared to consumer protection actions, Actions on third-party content mobilize more complex analyses, which can range from consideration and weighing of fundamental rights to intersection with criminal law (in cases where there is evidence of crimes against honor). This statement is corroborated by the opinion of Minister Luís Roberto Barroso, for whom Brazilian judges are trained to decide on specific cases, and not to “carry out systemic analyses” of the impact of their decisions in broader contexts (Estadão Conteúdo, 2024).

This question dialogues directly with one of the methodological limitations of the econometric study, which assumes the fixed cost of shares as a constant. That is, if

the tendency is for actions to be more complex than typical consumer actions, it is also reasonable to infer that the total value of the impact on the Judiciary presented here has a more conservative bias, and these values could be even higher.

Therefore, it is necessary to contextualize these costs in the reality of the Judiciary. Personnel and infrastructure expenses are necessary to maintain the active functioning of the system, which corresponded to 132.8 billion reais in 2023, equivalent to 1.2% of Brazilian GDP (CNJ, 2024). The percentage is considerably higher than that observed in other emerging countries and developed economies, where the average expenditure on the Judiciary is equivalent to 0.5% and 0.3% of GDP, respectively (Tesouro Nacional, 2023).

Thus, considering the high structural expenditure of the Judiciary, any budgetary increase resulting from new sources of litigation tends to intensify congestion and harm the quality of the service offered. As demonstrated above, the implementation of strict liability for platforms will lead to an increase in costs, which in the worst case scenario could reach up to 777 million reais for the Judiciary.

To get an idea of the magnitude of this impact in the highest scenario:

• The total cost can reach 485 times the cost of developing the artificial intelligence system, Victor, used to optimize the evaluation of REs by the STF, through the automatic return of resources that fall under one of the themes of general repercussion to the courts of origin (Teixeira, 2018).
• With the maximum amount, the Judiciary could pay for around 1,600 new federal judges for a period of one year (CJF, 2025) and 1,800 state judges (TJMG, 2025)5.

We reinforce that the increase in costs for the Judiciary should not be automatically condemned. In areas such as health, for example, judicialization is often used to guarantee access to treatments that are not yet incorporated into the SUS or are of high cost.

— functioning, in these cases, as a legitimate instrument for the enforcement of fundamental rights (INAFF, 2025).

What is at stake here, however, is another type of phenomenon: the possibility of creating incentives for the proliferation of private actions against platforms, in disputes between individuals, for damages caused by third parties. This type of litigation — often predatory — can not only generate high costs, but also compromise the effectiveness of Justice by shifting resources from legitimate cases to opportunistic disputes.

• The estimates were made based on the base table of subsidies available on the websites of the Federal Justice Council (CJF) and TJMG, therefore excluding additional charges and benefits, which are variable.

Abusive Litigation in Brazil

Predatory or abusive litigation consists of the use of the right of access to the Judiciary that deviates from or exceeds the limits of its social, legal, political or economic purpose, including in demands directed at specific defendants (CNJ, 2024). This practice is, today, one of the serious problems of the Brazilian Judiciary, causing great losses

and with a direct impact on the average processing time of processes (Minas Gerais Justice Intelligence Center, 2022).

According to a survey carried out by the Minas Gerais Justice Intelligence Center (CIJMG), the annual impact of predatory litigation is more than 10 billion reais (CIJMG, 2022)6. In the State of São Paulo, according to an estimate by the Center for Monitoring Demand Profiles of the Internal Affairs Department of Justice (NUMOPEDE), between 2016 and 2021, abusive litigation generated around 337 thousand cases per year (NUMOPEDE, 2023; Leitão, 2025), and a loss of around R$ 16.7 billion in the period and 2.7 billion per year to the treasury (NUMOPEDE, 2023).

The “abusive litigation industry” is a profitable business: in the State of Mato Erosso do Sul, according to a survey carried out by the TJMS Intelligence Center, between January 2015 and August 2021, 50,000 lawsuits were filed in banking matters. Among the total of these actions, around 80% (3G,704 cases) were presented by the same lawyer (TJMS Intelligence Center, 2022). In all these cases, it is estimated that there was an expense of R$148 million to the treasury, especially considering that 100% of the actions were sponsored by free justice (TJMS Intelligence Center, 2022).

Recognition of the so-called presumed moral damage, which eliminates the need to prove damage (Lins; Belmonte, 2024), is also a factor that encourages abusive litigation. Presumed moral damage (or “in re ipsa”) is one that does not require proof of damage, because the damage is considered a logical and automatic result of the fact. Known examples include improper registration in defaulter registers and lost luggage. In these cases, the Judiciary understands that the illicit act itself causes enough moral harm to justify compensation, and it is not necessary to prove the existence of material damage.

Until now, none of the votes presented expressly proposed the presumed moral damage – the votes of Toffoli, Fux and even Barroso do not provide clarity on how to prove the damage, whether in cases of objective or subjective liability. Therefore, the risk is jurisprudential: if local courts (especially in special courts) start to presume moral damage in actions with offensive content, fake news or false profiles, based on the logic of “undue exposure”, then we would have a scenario of presumed damage in practice, even if not in theory. This is not a theoretical risk: the jurisprudence of the STJ prior to the MCI, for example, recognized this situation of presumed damage in dozens of cases7.

• The estimate is based on a calculation of the average procedural cost, according to parameters established by the Institute for Applied Economic Research (Ipea), associated with a minimum incidence of 30% of civil cases with abuse of the right of action (CIJMG, 2022). The analysis considers two of the most recurring issues in State Courts in 2020: supplier liability/damage for moral compensation within the scope of consumer law and obligations/types of contracts (Vieira et al, 2023).
• For example, see AREsp 990096, REsp 1637026 and AREsp 166135, STJ.

It is necessary to recognize that the Brazilian Judiciary is aware of the issue of abusive litigation. In October 2024, the CNJ approved Recommendation No. 159/2024, which provides guidelines for judges so that they can identify and request measures in cases of abusive litigation (CNJ, 2024). In March 2025, the Special Court of the STJ decided that judges can demand, in a well-founded manner, and observing reasonableness

in each specific case, the amendment of the initial petition with documents that prove the interest to act and the authenticity of the postulation of the party suspected of promoting abusive litigation (STJ, 2025). The thesis was established under the repetitive appeal rite of Theme 1,198 and must be observed by courts across the country (STJ, 2025).

There is, therefore, an institutional mobilization to combat abusive litigation, with the aim of preserving legitimate access and the quality of an adequate service, and it makes sense that it is also observed – or at least considered – as part of the issue of the STF judgment.

Scenario Analysis and Assessment

Not every increase in costs for the Judiciary should be avoided: judicialization can be a legitimate way to correct abuses, combat rights violations and tackle serious problems such as misinformation and hate speech.

But legal costs aren’t the only factor at play either.. Studies have shown that models such as strict liability or notice and take down broad, although well-intentioned, can generate important side effects, such as increasing the excessive removal of content, or expanding the platforms’ discretionary power over what can and cannot circulate online (DG Connect, 2020). This is a field certainly marked by tensions – and the objective of this study is not to choose a side, but to contribute to a more informed debate in a specific area.

From the discussion of the results, our initial assessment is that combining (i) broad hypotheses of objective liability, (ii) low costs for filing the action, (iii) no need to make a prior complaint to the platform and (iv) possible interpretation of presumed moral damage, It seems like an explosive combination for abusive litigation, perhaps with factors not seen in any other sector in Brazil.

The construction of the scenarios designed in this study was guided by two central variables:

(i) the degree of accountability imposed on platforms and (ii) the speed of propagation of the effects of the decision on the Justice system. These two axes reflect, respectively, the normative content of the thesis to be established by the STF and the way in which this new legal framework can be interpreted and operationalized by different actors — legal operators, lawyers, judges, platforms and citizens.

Based on the discussion of the results, the econometric analysis and the evaluation of the literature on the topic, we seek here to rationalize the scenarios, associating each of them with possible regulatory contexts – even if (i) they do not occur naturally from the econometric model, but rather from the combination with secondary data and analysis of the content of the votes cast so far, and (ii) the theses can also fit into intermediate scenarios, that is, between each of the four discussed

These different scenarios are not presented in an exercise in “futurology”, but to qualify the regulatory debate: although they are preliminary projections that may even motivate future studies, depending on the normative content of the thesis established by the STF, different scenarios become more or less likely:

Scenario	Increase in shares in 5 years	Estimated cost of the increase (R$ million)	Rating and Comments

Scenario 1

(average increase, MCI rhythm)

158.403165,13

gradual growth with moderate impact

This scenario is more likely if the STF adopts a thesis of subjective responsibility (i.e. liability for failure to comply with a court decision or valid notifications), but with effects modulation. These effects can be modulated by tools such as limiting notice and take down hypotheses to specific hypotheses with low judicial discretion (e.g. graphic content, explicit violence), minimum criteria for the validity of a notification, or creating mechanisms for effectively measuring damage in cases of non-compliance, avoiding sudden increases motivated by abusive litigation. In this context, the growth of processes would be gradual, compatible with patterns observed after the MCI.

Scenario 2

(increase243,476average, peakMCI)250.88could occur, for example, if we have broad notice and take down hypotheses, there are no clear criteria for how notifications will be considered valid, and if the configuration of presumed moral damage is open to judicial discretion. As a result, even without a structural change in the liability regime, the incentives for abusive litigation would be high, resulting in an immediate but limited effect.gradual growth with high impactScenario 3This scenario is the most likely if the STF adopts a thesis of objective liability (that is, automatic liability, regardless of notification), but the impact of this decision on the volume of actions is dampened (upper threshold584,447,604,20by very short-term regulatory actions. These actions may include the insertion, in the thesis itself, of criteria such as minimum MCI rhythm requirements for characterizing damage or proof of a prior attempt at an extrajudicial solution; or even actions by the Legislature that create additional limits on civil liability. In this case, the impact would be high, but these mechanisms could stop the immediate effect of abusive litigation. This would create high and immediate incentives for mass litigation, including coordinated action by specialized companies, as has already been observed in other sectors.

immediate explosion with moderate impact

This scenario becomes likely if the STF adopt a thesis of subjective responsibility, but leave wide interpretative margins for judicial discretion and probable configuration of presumed moral damage. This scenario

In summary, this assessment presents the following scenarios:

If the line of votes from ministers Toffoli and Fux, which propose broad objective liability and with little delimitation of hypotheses, the risks of massive judicialization intensify, approaching Scenario 4, generating an impact of up to R$777 million;

On the other hand, if an intermediate thesis is articulated, like the one proposed by Minister Barroso — which maintains subjective responsibility for non-compliance with a court decision and creates hypotheses with specific criteria for notice and take down — the most likely scenario seems to be the Scenario 2, immediate explosion but moderate impact

(up to R$ 250 million of impact), but which can reduce Scenario 1 (up to R$ 165 million impact) according to the details of the thesis and the addressing of risks of abusive litigation.

Finally, if another minister presents a thesis based on objective liability, but with extensive details of rules for modulating effects, the cost should approach Scenario 3(up to R$604 million impact).

The main message of this study is that any change in the way digital platforms are held accountable must also take into account their impacts on the justice system. And, to make balanced decisions, it is essential to deepen the technical and data-based debate — helping the STF and Congress to minimize side effects without compromising the protection of rights.

POSSIBLE REGULATORY ALTERNATIVES

Although the main objective of this work consists of measuring the potential increase in costs resulting from the objective liability regime for digital platforms within the scope of the judgment of Topics 987 and 533 by the STF, it is relevant to present regulatory alternatives, which are not mutually exclusive and which have been proposed in recent months by some experts:

• Reversal of the Burden of Proof. For the professor at Fundação Getulio Vargas,

Osny da Silva Filho, the evidentiary difficulty is not a reason for the application of objective liability since, according to this author, the application of article 373, § 1 of the CPC/15, which allows the reversal of the burden of proof, in case of “impossibility” or “excessive difficulty” of producing evidence (Silva Filho, 2025).

• Exceptions Regime. An alternative that has been discussed consists of the systematization of new exceptions to article 19, expanding the list of article 21 of the MCI (Cruz, 2024). In this case, the STF, through the technique of interpretation in accordance with the Federal Constitution8 , could create exceptions to the rule of judicial notice and takedown. However, to avoid legal uncertainty and a scenario of trivializing excessive content removal, These hypotheses must be objective and clearly illegal.
• Systemic and Procedural Obligations. As already mentioned in the introduction to this work, the European Union’s Digital Services Act chose not to change the model of subjective responsibility already applied by Member States – the General Directorate of Communication Networks, Content and Technologies of the European Commission, in fact, rejected the adoption of the objective liability model due to the potential impact on freedom of expression and excessive removal of content by platforms (DG Connect, 2020). However, the DSA links the responsibility of providers to a systemic failure, that is, platforms are not held responsible for the mere existence of content, but for the lack of adoption of procedural obligations that enable its existence and non-removal on the platform.
• The defense of the Constitution-compliant interpretation of article 19 of the MCI was a position supported by some stakeholders in the case files of Themes 987 and 533, and gained strength from 2023 onwards. These actors propose interpretative adjustments to align the application of the norm with constitutional demands, considering that, in its literalness, the text can generate incompatibilities in specific contexts. Such positions imply an interpretative innovation, giving a meaning adapted to current realities.

For a more in-depth discussion on the topic of conformal interpretation, see SILVA, Virgílio Afonso da. Interpretation in accordance with the constitution: between triviality and judicial centralization. Journal of Constitutional and International Law, v. 2, no. 1, p. 191-210, Jan./Jun. 2006. This article served as a methodological reference for this section.

Conclusion

This study analyzed the economic impact of the STF on Themes 987 and 533, in which ministers will redefine the liability regime applicable to digital platforms. The research aimed to calculate the impact of the eventual adoption of objective liability rules for digital platforms on the public budget of the Brazilian Judiciary.

With the replacement of the current model by objective liability, an increase in the number of actions in the Judiciary is estimated, representing up to 2.6 times the current number, which would be equivalent to an increase of more than 750 thousand shares over a period of five years.

Although this volume may seem small compared to the more than 40 million annual cases in Brazilian Justice, These actions have a very different profile: They involve complex disputes, with consideration of rights, use of digital evidence, and may involve multiple instances. In other words, the problem here is not just volume — but of prioritization, structure and effectiveness of Justice as a whole, as the sudden growth of highly complex and specialized actions can divert the Judiciary’s attention from other urgent social issues.

In terms of costs, this increase is equivalent to at a value that can reach more than 770 million reais. In a scenario in which the Judiciary already represents a high cost to the State, any budgetary increase resulting from new sources of litigation tends not only to intensify congestion and harm the quality of the service offered, but also increase public pressure and criticism in relation to the costs of the Judiciary.

Understanding this impact is essential so that the judicial decision does not produce adverse side effects on the effectiveness of justice itself.. It is important that the review of the responsibility model provided for in article 19 of the MCI is carefully considered, considering both the constitutional foundations involved in the dispute and the systemic dimension of the decision on the health of the judicial system and its usefulness for society.

It is important to emphasize that the values indicated throughout this study correspond exclusively to the direct costs borne by the courts, not covering the total costs associated with litigation arising from undue liability of providers, such as litigation expenses borne by the parties. Furthermore, These values do not exhaust the broader social impacts of a possible increase in litigation caused by inadequate changes to the legal regime established by the MCI, such as the risk of impunity for users who, in fact, produce and disseminate illegal content and threats to freedom of expression – topics that also need to be explored.

More than a study on civil liability on the internet, this research also presents itself as an invitation for more decisions of general repercussion to consider in their impact not only the legal effects, but also the systemic costs for the Judiciary itself and its externalities. And we hope to continue contributing to this debate.

Suggestions for Future Studies

Based on the results and discussions of this work, as well as its methodological limitations, we highlight the following directions for new studies, which can continue and improve this research:

• Use of historical series: This study’s approach considered a fixed counterfactual scenario based on observed behavior, and does not capture how agents can change their behavior in the face of regulatory changes – the so-called “Lucas critique”. Future studies can use new models to simulate changes in content moderation and how this can increase or reduce the effects brought here, as well as explore other effects on social well-being.
• Indirect costs of judicialization for the justice system: the analysis focused on direct costs, and there are indirect costs that were not included and could have a relevant effect, such as delays in other areas of the Judiciary and impacts on precedents in other spheres (such as traditional media or consumer law).
• Qualitative analysis of perceptions among judges and legal practitioners: Future studies can investigate, through interviews and focus groups, how judges, defenders, lawyers and civil servants perceive the practical effects of objective liability, including risks of overload and impact on the quality of decisions.
• Effects for consumers and content creators: This study focused on the impacts on the Judiciary. Future research can analyze how

Different liability regimes affect internet users economically, considering compliance costs, barriers to entry and redistributive effects.

• International comparisons focusing on enforcement: Countries that have made relevant changes to their legislation on internet content moderation, such as Germany, can provide important inferences about effects in Brazil.
• Discussion of regulatory alternatives: although some distinct normative approaches have been presented, the in-depth and exhaustive analysis of these regimes was not covered by the main scope of this research. The detailed analysis of compatibility between regulatory arrangements and their practical feasibility of implementation in Brazil constitutes a promising line of research for future studies.
• Empirical effects of changing the liability regime on platform behavior: This study aims to predict, based on evidence, the budgetary and institutional costs of adopting the

objective liability for platforms, but does not propose to predict the possible behavioral effects arising from the implementation of this model for platforms. Future studies may propose to study these effects based on an economic analysis of law approach or comparative analyzes between countries with different liability regimes.

References

ABADIE, Alberto; DIAMOND, Alexis; HAINMUELLER, Jens. Synthetic Control Methods for Comparative Case Studies: estimating the effect of California’s tobacco control program. Journal of the American Statistical Association, [S.L.], v. 105, no. 490,

p. 493-505, jun. 2010. Informa UK Limited. http://dx.doi.org/10.11G8/jasa.200G. ap08746. Available at: https://economics.mit.edu/sites/default/files/publications/ Synthetic%20Control%20Methods.pdf. Accessed on: 07 Apr. 2024.

ABADIE, Alberto; GARDEAZABAL, Javier. The Economic Costs of Conflict: a case study of the Basque country. American Economic Review, [S.L.], v. 93,

n. 1, p. 113-132, 1 Feb. 2003. American Economic Association. http://dx.doi. org/10.1257/000282803321455188. Available at: https://economics.mit.edu/ sites/default/files/publications/The%20Economic%20Costs%20of%20Conflict.pdf. Accessed on: 07 Apr. 2025.

ABADIE, Alberto. Using Synthetic Controls: feasibility, data requirements, and methodological aspects. Journal Of Economic Literature, [S.L.], v. 59, no. 2, p. 391-425, 1 jun. 2021. American Economic Association. http://dx.doi.org/10.1257/jel.201G1450. Available at: https://www.aeaweb.org/content/file?id=1240G. Accessed on: 07 Apr. 2025.

ABEAR – BRAZILIAN ASSOCIATION OF AIRLINES. Since 2020, number

The number of lawsuits against airlines increased, on average, by 60% per year. 2024. Available at: https://www.abear.com.br/imprensa/agencia-abear/noticias/ since-2020-number-of-processes-against-airlines-has-increased-in-meday-60-a-year/. Accessed on: 03 Apr. 2025.

ABEAR – BRAZILIAN ASSOCIATION OF AIRLINES. Federal Court of-

sativa 37 “vulture sites”, which encourage excessive Judicialization in the airline sector. 2022. Available at: https://www.abear.com.br/imprensa/agencia-abear/ noticias/justica-federal-deactivates-37-vulture-sites-that-stimulate-the-excess-of-juicedicialization-in-the-airline-sector/. Accessed on: 03 Apr. 2025.

ARTICLE 1G.Islamic Republic of Iran: Computer Crimes Law. 2012. Available at: https://www.article1G.org/data/files/medialibrary/2G21/12-01-30-FINAL-iran-WE-B%5B4%5D.pdf. Accessed on: 13 Mar. 2025.

BATISTA, Rafael. The future of digital platforms in Brazil. 2024. Available at: https://www.jota.info/artigos/o-futuro-das-plataformas-digitais-no-brasil. Accessed on: 26 Mar. 2025.

BRAZIL. Superior Court of Justice. Special Appeal No. 1986323. Brasília, DF, September 6, 2022. Electronic Justice Diary. [S.I], 13 September. 2022. Available at: https://scon.stj.jus.br/SCON/pesquisar.jsp?pesquisaAmigavel=+2021%-2F0303507-3&b=ACOR&tp=T&numDocsPagina=10&i=1&O=&ref=&processo=&e-menta=¬a=&filtroPorNota=&orgao=&rapporteur=&uf=&classe=&juizo=&data=&dtpb=&d-tde=&operator=e&thesaurus=JURIDICO&p=true&livre=2021%2F0303507-3. Accessed on: 04 Apr. 2025.

BRAZIL. Recommendation No. 159 of 10/23/2024. [S.I], Available at: https://atos. cnj.jus.br/atos/detalhar/5822. Accessed on: 25 Mar. 2025.

CATTANEO, Matias D.; FENG, Yingjie; PALOMBA, Filippo; TITIUNIK, Rocio. Scpi: uncertainty quantification for synthetic control methods. Arxiv, [S.L.], p. 1-41, 2022. ArXiv. http://dx.doi.org/10.48550/ARXIV.2202.05G84. Available at: https://arxiv.org/ abs/2202.05G84. Accessed on: 07 Apr. 2025.

Minas Gerais Justice Intelligence Center. CIJMG Technical Note nº 01/2022. 2022. Available at: https://www.tjmg.jus.br/data/files/4G/80/E5/70/ DF212810B8EE0B185ECB08A8/NT_01_2022%20_1_%20_1_.pdf. Accessed on: 04

apr. 2025.

Intelligence Center of the Mato Erosso do Sul Court of Justice. Technical Note 01/2022. 2022. Available at: https://www.tjms.jus.br/storage/cms-arquivos/62a-318e6cbe701Gb873fa0a4d8d585GG.pdf. Accessed on: 25 Mar. 2025.

CHOI, Hyunyoung; VARIAN, Hal. Predicting the Present with Google Trends. Economic Record, [S.L.], v. 88, no. 1, p. 2-e, jun. 2012. Wiley. http://dx.doi.org/10.1111/j.

1475-4932.2012.0080G.x. Available at: https://papers.ssrn.com/sol3/papers. cfm?abstract_id

CJF – Federal Justice Council.Remuneration table for federal judges. 2025. Available at: https://www.cjf.jus.br/cjf/transparencia-publica-1/ personnel-information/remuneration-structure/2025/remuneration-table-

-of-servers. Accessed on: 04 Apr. 2025.

National Council of Justice. Justice in Numbers 2024. Brasília: National Council of Justice, 2024. Available at: https://www.cnj.jus.br/wp-content/uplo-ads/2025/02/justica-em-numeros-2024.pdf. Accessed on: 24 Mar. 2025.

CRUZ, Francisco Brito. Changes on the internet require reinterpretation of the Marco Civil by the STF. Folha de S. Paulo. São Paulo. 19 Sep. 2024. Available at: https:// www1.folha.uol.com.br/ilustrissima/2024/0G/mudancas-na-internet-demandam-

-reinterpretation-of-the-civil-marque-by-stf.shtml. Accessed on: 31 Mar. 2025.

Directorate-General for Communications Networks, Content and Technology. Digital Services Act: Questions and Answers. 2024. Available at: https:// digital-strategy.ec.europa.eu/en/faqs/digital-services-act-questions-and-

-answers. Accessed on: March 21, 2025.

Directorate-General for Communications Networks, Content and Technology. Impact Assessment of the Digital Services Act. 2020. Available at: https:// digital-strategy.ec.europa.eu/en/library/impact-assessment-digital-services-act. Accessed on: March 21, 2025.

STATE CONTENT. Barroso: The Judiciary in Brazil has a different role than in the rest of the world. 2024. Available at: https://noticias.uol.com.br/

latest-noticias/agencia-state/2024/11/21/barroso-judiciario-no-brasil-tem-

-role-different-than-in-the-rest-of-the-world.htm. Accessed on: 27 Mar. 2025.

FIRO, Sergio, and POSSEBOM, Vitor. “Synthetic Control Method: Inference, Sensitivity Analysis and Confidence Sets” Journal of Causal Inference, vol. 6, no. 2, 2018, https://doi.org/10.1515/jci-2016-0026. Accessed on: 27 Mar. 2025.

GERCINA, Cristiane. Labor Court receives 2 million new lawsuits in 2024, a record after labor reform. Folha de S. Paulo. São Paulo, Feb.

2025. Available at: https://www1.folha.uol.com.br/mercado/2025/02/justica-

-do-labor-receives-2-million-new-shares-in-2024-record-after-reinforcement-ma-trabalhista.shtml. Accessed on: 03 Apr. 2025.

GIMENES, José Jácomo. Republic of proceedings, an ongoing national tragedy. 2024. Available at: https://www.jota.info/artigos/republica-de-

-processes-a-national-tragedy-in-progress. Accessed on: 26 Mar. 2025.

INAFF. Number of judicialization in health registers increase. 2025. Available at: https://www.inaff.org.br/noticias/nume-

ro-de-judicializacao-na-saude-registra-aumento/#:~:text=Os%20da-dos%2C%20que%20s%C3%A3o%20do,16%25%20entre%202022%20e%20

2023. Accessed on: 24 Mar. 2025.

INGIZZA, Carolina; CARVALHO, Mirielle. Why the number of lawsuits against airlines is so high in Brazil. 2024. Available at: https:// www.jota.info/justica/por-que-o-numero-de-processos-contra-companias-

-aereas-e-tao-alto-no-brazil. Accessed on: 25 Mar. 2025.

LEITÃO, Matheus. Predatory litigation affects the economy and business, says expert. Available at: https://veja.abril.com.br/coluna/matheus-lei-tao/a-litigancia-predatoria-afeta-economia-e-os-negocios-diz-especialista#-google_vignette. Accessed on: 25 Mar. 2025.

LINS, Julia Vieira de Castro; BELMONTE, Renata Martins. Predatory litigation and the cost of judicialization in the Brazilian airline sector. 2024. Available at: https://www.jota.info/artigos/litigancia-predatoria-e-o-custo-da-judicializacao-

-in-the-Brazilian-aero-sector. Accessed on: 25 Mar. 2025.

LUCAS, Robert E.. Econometric policy evaluation: a critique. Carnegie-Rochester Conference Series On Public Policy, [S.L.], v.1, p. 19-46, Jan. 1976. Elsevier BV. http://dx.doi.org/10.1016/s0167-2231(76)80003-6. Available at: https://www.sciencedirect.com/science/article/abs/pii/S0167223176800036. Accessed on: 07 Apr. 2025.

MELLO, Patrícia Campos. Countries discuss holding bigtechs responsible, but Brazil risks being “jabuticaba”.Folha de S. Paulo. São Paulo. 08 Dec. 2024. Available at: https://www1.folha.uol.com.br/poder/2024/12/paises-discutem-

-more-responsibility-for-big-techs-but-brazil-risks-being-jabuticaba. shtml. Accessed on: 04 Apr. 2025.

MIRAGE, Bruno. Civil Liability – 29 Edition 2021. 2nd ed. Rio de Janeiro: Forense, 2021. E-book. p.67. ISBN 9788530994228. Available at: https:// integra.minhabiblioteca.com.br/reader/books/G788530GG4228/. Accessed on: 12 Mar. 2025.

MOYNIHAN, Harriet; PATEL, Champa. Restrictions on online freedom of expression in China: the domestic, regional and international implications of china’s policies and practices. The domestic, regional and international implications of China’s policies and practices. 2021. Available at: https://www. chathamhouse.org/sites/default/files/2021-03/2021-03-17-restrictions-online-

-freedom-expression-china-moynihan-patel.pdf. Accessed on: 13 Mar. 2025.

MURTA GOYANES LAWYERS. Manifestation – MercadoLivre.com. Available in: https://redir.stf.jus.br/estfvisualizadorpub/jsp/consultarprocessoele-tronico/ConsultarProcessoEletronico.jsf?seqobjetoincidente=516054G. Accessed on: 07 Apr. 2025.

NUMOPEDE – Demand Profile Monitoring Center. Recognition of Predatory Litigation. 2023’. Available at: https://api.tjsp.jus.br/ Handlers/Handler/FileFetch.ashx?codigo=151470. Accessed on: 25 Mar. 2025.

RAMALHO, Renan. Censorship proposed by Toffoli favors compensation industry against social networks. Eazeta of the People. Curitiba. 10 Dec. 2024. Available at: https://www.gazetadopovo.com.br/vida-e-cidadania/censura-proposta-

-by-toffoli-favorece-indemnity-industry-against-social-networks/. Accessed on: 03 Apr. 2025.

RAMOS, P. H. et al. Interests at stake in the STF: Mapping the Dispute over the Internet Civil Framework. Radar Reglab, no. 1. São Paulo: Reglab, 2024.

REIS, Nadja Alves dos. Judicialization: in the STF, minister Nísia states: “this is a milestone for Brazil and the SUS”. 2024. Available at: https://www. gov.br/saude/pt-br/assuntos/noticias/2024/outubro/judicializacao-no-stf-mi-nistra-nisia-affirms-this-is-a-milestone-for-brazil-and-sus. Accessed on: 18 Mar. 2025.

RESENDE, Guilherme Mendes (ed.). Digital Platform Markets (REVISED AND UPDATED EDITION). Cade’s notebooks, [S.L.], p. 1-278, 2023. Cade.

http://dx.doi.org/10.528G6/dee.cc2.023. Available at: https://cdn.cade.gov. br/Portal/centrais-de-conteudo/publicacoes/estudos-economicos/cadernos-

-do-cade/Caderno_Plataformas-Digitais_Atualizado_2G.08.pdf. Accessed on: 02 Apr. 2025.

SILVA FILHO, Osny da.Four precautions for judging article 1G. 2025. Available at: https://portal.fgv.br/artigos/quatro-cautelas-para-o-julga-ment-of-article-1G. Accessed on: 01 Apr. 2025.

Superior Court of Justice. Special Court repeatedly decides that judge can demand documents to prevent abusive litigation. 2025. Available at: https://www.stj.jus.br/sites/portalp/Paginas/Comunicacao/Noticias/2025/ 20032025-Special-Court-decides-in-repetitive-that-judge-can-require-documents-to-curb-abusive-litigation.aspx#:~:text=A%20litig%C3%A2ncia%20 abusive%20occurs%2C%20por,see%20livre%20dessa%20press%C3%A3o%20 procedural. Accessed on: 25 Mar. 2025.

Federal Supreme Court. STF overturns Labor Reform rules that restricted free access to Labor Court. 2021. Available at: https://portal.stf.jus.br/noticias/verNoticiaDetalhe.asp?idConteudo=47515G&o-ri=1. Accessed on: 04 Apr. 2025.

TEFFÉ, Chiara Spadaccini de. Content moderation and civil liability on digital platforms: a current look. 2024. Available at: https:// itsrio.org/pt/publicacoes/moderacao-de-conteudo-e-responsibility-civil-

-on-digital-platforms-a-current-look-2/. Accessed on: 04 Apr. 2025.

TEFFÉ, Chiara Spadaccini de; SOUZA, Carlos Affonso.Content moderation and civil liability on digital platforms: a look at the Brazilian, American and European experiences. In: Joyceane Bezerra de Menezes, Fernanda Nunes Barbosa. (Org.). The priority of the human person in Civil-Constitutional Law: studies in honor of Maria Celina Bodin de Moraes.

1.ed. Focus, 2024, p. 25-37.

TEIXEIRA, Matheus. STF invests in artificial intelligence to speed up processes. 2018. Available at: https://www.jota.info/coberturas-es-specials/inova-e-acao/stf-aposta-inteligencia-artificial-celeridade-processos. Accessed on: 03 Apr. 2025.

National Treasury. Expenditure by General Government Function – 2023 Results: cofog classification ⠳ classification of functions of government. COFOG Classification – Classification of Functions of Government. 2025. Available at: https://tesourotransparente.gov.br/publicacoes/cofog-despesas-por-funcao-

-do-governo-general/2023/114. Accessed on: 07 Apr. 2025.

TJMG – Court of Justice of the State of Minas Gerais. Annex III – C) Remuneration structure – Members of the Judiciary. 2025. Available at: https:// www8.tjmg.jus.br/transparencia/relatorios/mãesMagistratura.jsf. Accessed on: 04 Apr. 2025.

VIEIRA, Mônica Silveira; SANTOS, Daniel Eeraldo Oliveira; ASSUNAÇÃO, Rafaella Costa da Rocha.Predatory litigation consumes more than R$25 billion from public coffers annually. 2023. Available at: https://www.conjur.com. br/2023-out-31/opiniao-litigancia-predatoria-consome-25-bilhoes/. Accessed on: 04 Apr. 2025.

WANG, Jufang. Platform Responsibility with Chinese Characteristics. 2022. Available at: https://digitalplanet.tufts.edu/wp-content/uploads/2023/02/

DD-Report_1-Jufang-Wang-11.30.22.pdf. Accessed on: 13 Mar. 2025.

Reglab Methodology Annex

POLICE BRIEF

AUTHORITY: Pedro Henrique Ramos, João Ricardo Costa Filho and Daniela Naomi Shimabukuro Nomura

Title	The Price of Moderation: Impacts on the Judiciary and the Debate on the Review of the Internet Civil Framework by the STF
Research Question	What is the projected economic impact of the STF decision in Topic 987 on the Brazilian Judiciary, considering the expected increase in the number of lawsuits against digital platforms?
Methodology Summary	Synthetic control model to estimate the counterfactual effect of the STF decision on Topic 987 in relation to the projection of the number of cases in state courts of justice selected based on ARIMA models. Calculation of the present value of the total cost of increasing processes in four scenarios.
Data Collection	Primary data: Collection of data relating to searches on Google Trends and share values ​​and trading volume carried out between March 11 and 14, 2025.<br>At the same time, the number of processes in the selected TJs were collected and added together to compose the model’s variable of interest. In addition, the TJSP budget was verified in the state government’s LOA, as well as the number of first-instance court orders in 2024, which complemented the database used in the empirical analysis. Secondary data: In parallel, we carried out extensive documentary research to contextualize the results found in the synthetic control analysis. Academic articles, news, opinion articles, official documents published by public bodies were consulted (for example, the CNJ’s annual Justice in Numbers report), as well as reports from organizations (for example, Chatam House<br>/ Article 19 reports on content moderation and freedom of expression). These materials serve to contextualize the data resulting from the econometric analysis.
Data Analysis	A counterfactual scenario was designed using a fictitious company (called “Company S”). Initially, data from companies such as Facebook/Meta, Twitter/X, Google (Youtube) were considered, but that of the last two was discarded, focusing on the data available from FacebookG. The approach adopts a logic analogous to that of randomized controlled trials: a control group (which would be impacted by the MCI) and a treatment group (the treatment here is defined as the implementation of article 19 of the MCI). In the case of the synthetic control model, the difference is in the fact that the control sample is not just collected, but synthetically constructed. Actions from the following Courts of Justice were considered: TJSP, TJMG, TJRJ, TJBA, TJPR and TJRS.<br>For the creation of Company S, the following companies were considered: Bradesco, Itaú, Santander, Latam, Magazine Luiza, Amazon, Lojas Renner and Casas Bahia. We tried to use companies in the control group that were not directly influenced by the MCI, but that had contact with the end consumer. In constructing the synthetic control group, in addition to the number of processes of these companies, it was also necessary to use some of their characteristics. To achieve this, in this work, the following were considered: searches on Google (Google Trends), the value of shares in dollars and the volume of shares traded. All empirical analysis data is annual.

• Facebook Brazil processes were used as an empirical reference due to the availability of existing data for analysis. In the case of Twitter/X, there was not enough history and volume for the econometric analysis to be carried out. Other companies in the sector (e.g. TikTok and Kwai/Joyo) were not present in Brazil at the time of the MCI. However, the volume of processes of these companies were considered in the subsequent stage (total volume calculation). In the case of Google, Orkut stopped operating in the country a few months after the Marco Civil and, in the case of Youtube, the characteristics of the processes that involved removing content undermined the credibility of the analysis, due to the significant number of processes that dealt not with article 19, but with article 21 (removal of copyright content), which is why its analysis had to be excluded from the final results.

Data Analysis

Data on Google Trends can anticipate economic indicators (Choi; Varian, 2012) and can be useful in revealing agents’ decisions that are often not directly observed. In the case of the present study, the combination, in lowercase letters, “company name” and “auditor” was used (for example, “facebook Ouviria” or “Bradesco Ouviria”). In the data collection stage, only searches were carried out in Brazil. O The objective of including this variable is exploratory and aims to capture the feeling of dissatisfaction with the company that could motivate the initiation of legal proceedings, even if it cannot be used as an index of direct causality for formal litigation. Google Trends produces an indicator that goes from zero to 100, in which the highest value is assigned to the period with the highest number of searches for the expression under analysis. As the data in this study are annual, the monthly average for the year was considered for each company. The inclusion of the variables “value of shares in dollars” and “volume of shares traded” hypothesizes that, possibly, the motivations for legal proceedings may be related to the market value of companies (as a proxy for economic-financial performance) and the interest of investors in the company (approximated here by the volume of shares traded). The data was obtained from Yahoo Finance. The closing value for the year and the volume of transactions associated with this value were considered. Once the effect of the MCI was estimated in the five years following its introduction, it was assumed that the impact on companies operating in the sector today would be obtained through the percentage difference between the control group and the treatment group. Thus, it was assumed that the companies Bytedance Brasil, Facebook Brasil, Joyo Tecnologia Brasil and From this, some scenarios were constructed to be detailed in the Main Results section. After estimating the change in the number of cases, the average cost per action was calculated, making it possible to estimate the increase in the Judiciary’s budget resulting from the STF decision. To estimate the average unit cost of an action, the TJSP budget was used, which appears in the 2024 Annual Budget Law, divided by the number of first-degree orders in court in 2024. The result is an estimate of R$1,263.78 per action. When constructing each scenario for the impact on the public budget of the STF decision, the present value (for the period 2025-2029; at January 2025 prices) of the increase in the budget of the Judiciary. The total cost simulations were carried out in real terms (assuming, therefore, that the average unit cost per share will remain constant). Thus, the discount rate must be a real rate. It was calculated based on median projections for the inflation rate (measured by IPCA) and the Selic rate from the Top 5 long-term group of the Focus Report of the Central Bank of Brazil on 3/21/2025.

Bias Reduction Procedures

To reduce bias, empirical analysis references widely consolidated in the literature were used. Furthermore, the methodological approach was discussed and evaluated internally on two occasions so that suggestions and criticisms could be incorporated into the work before the analysis was carried out. Other procedures adopted include:<br>Double Validation in Critical Stages: for the “Discussion of Results” and “Analysis and Evaluation of Scenarios” section, the three researchers reviewed the text independently. In cases of disagreement, a fourth person was called to arbitrate and reach consensus.<br>Recording and Methodological Transparency: we kept detailed records of all versions of files and research, preserving history and allowing for a more systematic review.

Other Methodological Limitations	Hypothesis that the MCI counterfactual would be replicated: the construction of the scenarios presented in this study had as its main hypothesis that the counterfactual behavior of the number of actions in relation to the MCI would be emulated after the judgment of Topics 987 and 533 by the STF. Lack of a structural model: to overcome the above limitation, a structural model could be created that considered how agents would respond in relation to the possible scenarios delimited in this study (and other scenarios not addressed here). Reliance on Open Access Sources: The study relied significantly on research carried out in open access databases, news portals, newspapers and academic journals. Dependence on these sources may restrict the scope of the analysis, considering that relevant materials present in restricted access or specialized databases may not have been considered, which may compromise the completeness and depth of the text presented.	Hypothesis that the MCI counterfactual would be replicated: the construction of the scenarios presented in this study had as its main hypothesis that the counterfactual behavior of the number of actions in relation to the MCI would be emulated after the judgment of Topics 987 and 533 by the STF. Lack of a structural model: to overcome the above limitation, a structural model could be created that considered how agents would respond in relation to the possible scenarios delimited in this study (and other scenarios not addressed here). Reliance on Open Access Sources: The study relied significantly on research carried out in open access databases, news portals, newspapers and academic journals. Dependence on these sources may restrict the scope of the analysis, considering that relevant materials present in restricted access or specialized databases may not have been considered, which may compromise the completeness and depth of the text presented.	Hypothesis that the MCI counterfactual would be replicated: the construction of the scenarios presented in this study had as its main hypothesis that the counterfactual behavior of the number of actions in relation to the MCI would be emulated after the judgment of Topics 987 and 533 by the STF. Lack of a structural model: to overcome the above limitation, a structural model could be created that considered how agents would respond in relation to the possible scenarios delimited in this study (and other scenarios not addressed here). Reliance on Open Access Sources: The study relied significantly on research carried out in open access databases, news portals, newspapers and academic journals. Dependence on these sources may restrict the scope of the analysis, considering that relevant materials present in restricted access or specialized databases may not have been considered, which may compromise the completeness and depth of the text presented.
Use of Software		Software<br><br><br>MS Office Suite	Use in Research<br><br><br>Editing text, spreadsheets and graphs
Use of Software		Adobe C Suite	Layout and finalization of graphics and illustrations
Use of Software		ChatGPT	Grammar review (spelling, grammar, search for synonyms), language adaptation and adjustments to pre-prepared excerpts
Use of Software		More UFSC	Generation of bibliographic references in the ABNT model
Use of Software		R	Synthetic control model estimation
Use of Software
Ethical Guidelines	This research was financed by Google Brasil Internet Ltda., a company that has a direct interest in the result of Theme 987 of the STF. To ensure the integrity of this work, the authors developed, conducted and analyzed the study independently, without any contribution or interference from the company, which also did not influence or interfere in the interpretation of the results.<br>The authors maintain full professional independence and responsibility for the content and conclusions of this work.<br>Respect for Privacy and Confidentiality: The data used is in the public domain and was obtained from accessible sources, without violating the privacy or confidentiality of any individual or institution.<br>Responsible Use of Data Public: Although the data analyzed is public, its use was made in a responsible and ethical manner, with the exclusive objective of academic research.<br>Non-discrimination and Respect for Diversity. The research was conducted in a way that respects diversity and avoids any form of discrimination.	This research was financed by Google Brasil Internet Ltda., a company that has a direct interest in the result of Theme 987 of the STF. To ensure the integrity of this work, the authors developed, conducted and analyzed the study independently, without any contribution or interference from the company, which also did not influence or interfere in the interpretation of the results.<br>The authors maintain full professional independence and responsibility for the content and conclusions of this work.<br>Respect for Privacy and Confidentiality: The data used is in the public domain and was obtained from accessible sources, without violating the privacy or confidentiality of any individual or institution.<br>Responsible Use of Data Public: Although the data analyzed is public, its use was made in a responsible and ethical manner, with the exclusive objective of academic research.<br>Non-discrimination and Respect for Diversity. The research was conducted in a way that respects diversity and avoids any form of discrimination.	This research was financed by Google Brasil Internet Ltda., a company that has a direct interest in the result of Theme 987 of the STF. To ensure the integrity of this work, the authors developed, conducted and analyzed the study independently, without any contribution or interference from the company, which also did not influence or interfere in the interpretation of the results.<br>The authors maintain full professional independence and responsibility for the content and conclusions of this work.<br>Respect for Privacy and Confidentiality: The data used is in the public domain and was obtained from accessible sources, without violating the privacy or confidentiality of any individual or institution.<br>Responsible Use of Data Public: Although the data analyzed is public, its use was made in a responsible and ethical manner, with the exclusive objective of academic research.<br>Non-discrimination and Respect for Diversity. The research was conducted in a way that respects diversity and avoids any form of discrimination.