Who Watches Whom When Power Finds Few Limits? Brazil 2026 Presidential Election

Executive, Congress, Judiciary, and a proposal for the State to recognize patterns across decades without creating a new unchecked power

Maybe you have felt this before.

An ordinary person misses a deadline.

Pays a fine.

Has an account blocked.

Quickly discovers that not knowing the rule does not erase its consequences.

Then they open the news and find cases involving millions or billions moving for years through appeals, jurisdictional disputes, annulments, and statutes of limitation.

That alone does not prove favoritism or illegality.

But it produces a feeling.

And perhaps we can begin precisely there:

Does the same law reach everyone in the same way?

The law may be the same. The paths through it are not.

Imagine two Body-Territories facing the same system.

One has little money, little time, and limited capacity to sustain a legal dispute for years.

The other can hire highly specialized teams, reorganize corporate structures, produce dozens of appeals, and follow cases for decades.

Formally, the law remains the same.

Materially, the available movements are different.

Brazilian researchers have already analyzed the advantages of so-called repeat players: actors who, through repeated experience, resources, and specialization, learn to navigate the legal system more effectively.

This does not mean that those with money buy judicial decisions.

It means recognizing something simpler:

Economic resources also expand legal possibilities.

Within the Body-Territory, a rule never reaches an abstract citizen.

It reaches income.

Time.

Relationships.

Knowledge.

Territory.

And different possibilities for defense.

Perhaps a Decolonial Neuroscience must also consider what happens when a society repeatedly experiences the feeling that small infractions receive rapid consequences while large economic structures seem able to move through much longer labyrinths.

Institutional trust is also produced through experience.

Three branches of power should create limits. But who limits all three?

The Executive needs power to govern.

It appoints ministers.

Manages institutions.

Executes the budget.

Negotiates with Congress.

Without these capacities, an elected president cannot transform a political program into action.

But that capacity also requires oversight.

Appointments, contracts, and public resources can create conflicts of interest regardless of who is in government.

The Legislative branch must supervise the Executive.

And over recent decades, it has acquired enormous influence over the federal budget.

Raul Bonfim, Joyce Hellen Luz, and Vitor Vasquez demonstrated in 2023 that mandatory parliamentary amendments effectively changed the former dominance of the Executive over part of Brazil's budgetary cycle. (brazilianpoliticalsciencereview.org)

This can represent decentralization.

But decentralizing power does not automatically democratize it.

We need to know:

Who requested the money?

Who received it?

How much did it cost?

Who executed the project?

What was the result?

In 2026, Brazil's Federal Court of Accounts continued to identify traceability problems in parliamentary amendments and recommended mechanisms capable of following resources from their initial designation through final execution. (portal.tcu.gov.br)

Then we reach the Judiciary.

Its independence is indispensable.

A judge who can be removed simply for contradicting a president is not independent.

But independence also cannot mean:

No one may ask how decisions, relationships, and institutional patterns were formed.

And so we return to the central question:

Who controls those who control?

Perhaps the problem lies in the "engine room"

Argentine constitutional scholar Roberto Gargarella argues that Latin America significantly expanded social and constitutional rights while frequently preserving older structures of concentrated power in what he calls the Constitution's "engine room."

We changed rights.

But perhaps we did not transform, with the same intensity, the mechanisms through which power is produced and controlled. (researchgate.net)

Colombian researcher Sara Piedrahita Sierra adds another dimension through her analysis of institutional capture: economic elites can convert private resources into greater capacity to influence political organizations and public institutions. (revistas.upb.edu.co)

This does not mean that every relationship among businesspeople, politicians, lawyers, or judges is improper.

That would be precisely the mistake we want to avoid.

The question is different:

Can the State recognize when certain relationships and modes of operation begin to repeat?

What if the State could remember?

Here we introduce a BrainLatam proposal.

A case ends.

Years later, another begins.

The corporate registration number changes.

The holding company changes.

The fund changes.

The contract changes.

The representatives change.

But some individuals, ultimate beneficial owners, intermediaries, or structures may reappear.

Could one person track all of these relationships over thirty years?

Probably not.

Institutional intelligence could.

We propose a Public Center for Procedural and Patrimonial Intelligence, operated with governmental AI and subjected to independent institutional oversight.

It would not be an AI judge.

It would not convict.

It would not determine guilt.

Its first function would be:

to allow the State to have memory.

One year or R$1 million: a second layer of intelligence

As a starting point for debate, we propose two triggers.

Cases involving:

more than twelve months of proceedings

or

an amount equal to or greater than R$1 million

would automatically enter an additional layer of analysis.

The thresholds could later be recalibrated by law and empirical evidence.

Entering this system would not mean that anyone had become a suspect.

It would simply increase institutional capacity to recognize complex relationships.

Within legally accessible databases, the AI could construct historical graphs involving:

• individuals and corporate entities;

• ultimate beneficial owners;

• funds and holding companies;

• contracts and public procurement;

• parliamentary amendments;

• corporate ownership;

• public appointments;

• previous judicial and administrative proceedings;

• decisions by oversight bodies;

• legally accessible asset changes;

• recurring relationships among public and private actors.

We would stop looking only at cases.

We would begin to see patterns over time.

Synchronizing with COAF

Brazil's Financial Intelligence Unit, COAF, already offers a clue as to how this might work.

Its 2026 regulation strengthened due diligence concerning ultimate beneficial owners, corporate structures, source of funds, continuous monitoring, and traceable documentation of analyses. (gov.br)

Our proposal is that the Center should be able to legally interoperate with this financial intelligence.

This would not mean "unrestricted access to bank accounts."

Information sharing would need to respect institutional competences, confidentiality, and formal procedures.

Brazil's Supreme Federal Court itself allows Financial Intelligence Reports to be shared under specific conditions, while requiring confidentiality, formal channels, and the possibility of subsequent judicial review. (portal.stf.jus.br)

The AI would receive signals.

Compare relationships.

Detect recurrences.

And produce intelligence alerts.

Never convictions.

And synchronize social networks as well?

Yes, but here we need to distinguish between two very different things.

What a person has made public can form part of an open-source intelligence layer.

Publicly disclosed corporate participation.

Public events.

Public photographs.

Posts.

Declared professional relationships.

Publicly presented institutional relationships.

All of these may help us understand a network.

Private messages, restricted location data, protected content, and other confidential communications should not be automatically searched simply because someone appears in a legal proceeding.

They must remain subject to applicable legal safeguards and, when necessary, judicial authorization.

Because a photograph does not prove corruption.

A friendship does not prove a crime.

Being someone's lawyer does not make a person part of their conduct.

Correlation cannot become conviction.

But recurring patterns can produce a question for human investigators:

Is there a relationship here that deserves to be understood?

What if the same intelligence also observed the State itself?

Perhaps this is the most important part.

The system should not observe only businesspeople or private citizens.

It should also be able to detect unusual patterns involving members of the:

Executive.

Legislative.

Judiciary.

Oversight institutions.

Across decades, it could compare legally available information about contracts, rulings, parliamentary amendments, appointments, corporate relationships, beneficiaries, and conflicts of interest.

Not to create a "secret criminality score."

And certainly not to claim that a judge, legislator, or president is a criminal because an algorithm found correlations.

The objective would be to identify unusual risk patterns significant enough to justify independent human auditing.

If certain people, companies, funds, decisions, and transfers repeatedly appear together over twenty years, perhaps the State should be capable of noticing.

The case may expire. Institutional memory does not need to disappear.

Here we make a fundamental distinction.

We are not proposing that illegally obtained evidence become valid.

Nor are we proposing that an AI automatically eliminate statutes of limitation established by law.

The right to challenge evidence remains.

The right to defense remains.

The requirement that evidence be lawfully obtained remains.

What we propose is separating:

the expiration of the possibility to punish

from

the preservation of legally obtained institutional memory.

Perhaps we can summarize it this way:

Justice may recognize that a particular event can no longer result in punishment. The State does not therefore need to learn to forget it.

Previously and lawfully preserved information could help a future investigation recognize operating patterns, recurring beneficiaries, or structures that would remain invisible if analyzed only in isolation.

And Lula? What does his plan propose about this struggle over power?

Here we need to include Lula with the same care applied to the other candidates.

His 2026 government program does not place a structural reform of the Supreme Federal Court at the center in the same way as proposals from candidates such as Flávio Bolsonaro, Zema, or Cury.

His institutional concern appears primarily elsewhere:

in the relationship among the Executive, Congress, and the federal budget.

The program registered with the Superior Electoral Court criticizes the major expansion of parliamentary amendments and argues that mandatory amendments and mechanisms associated with the so-called secret budget have reduced the Executive's capacity to coordinate public spending and fragmented the allocation of resources.

The proposal includes revisiting this system and expanding participatory budgeting mechanisms.

This places Lula before a legitimate question:

How much budgetary power needs to remain with an elected president for that president to actually execute the program chosen by voters?

But the inverse question also exists:

How do we prevent the recovery of Executive capacity from simply reconcentrating power in the Presidency?

This point matters because Lula's own government is also subject to oversight.

When reviewing the presidential accounts for 2025, Brazil's Federal Court of Accounts approved them with reservations and pointed, among other issues, to deficiencies in transparency and traceability in certain expenditures involving parliamentary amendments classified within discretionary Executive programs. (portal.tcu.gov.br)

At the same time, the current government maintains a 2025–2027 Integrity and Anti-Corruption Plan, coordinated by the Office of the Comptroller General, which includes institutional strengthening, greater budget transparency, including parliamentary amendments, integrity in public-private relations, and improvements in the detection and accountability of unlawful conduct. (gov.br)

Both pieces of information need to exist together.

The program promises more control and transparency.

The oversight institutions themselves continue to identify areas that need correction.

This is how our analysis should work:

No government receives immunity simply because it proposes to supervise others.

And the other candidates?

Flávio Bolsonaro proposes reducing the scope of individual Supreme Court decisions, reviewing aspects of the Court's jurisdiction, establishing restrictions involving relatives of judges, and ending presidential re-election.

Romeu Zema also places institutional limitations on the Supreme Federal Court among his reform proposals.

Ronaldo Caiado combines ending re-election with greater traceability of parliamentary amendments and criticism of concentrated power both in Congress and the Judiciary.

Augusto Cury proposes semi-presidentialism and eight-year terms for Supreme Court justices.

Edmilson Costa proposes a deep reorganization of the Judiciary and a unicameral Parliament.

Rui Costa Pimenta goes further and proposes abolishing the Supreme Federal Court in its current form.

Renan Santos emphasizes administrative and federal reforms.

Clariana Barão focuses on results-oriented federalism, data, targets, and evaluation.

The official programs registered with the Superior Electoral Court make it possible to compare these different institutional architectures. (tse.jus.br)

But perhaps our proposal introduces a question that cuts across all of them:

It is not only about who gains or loses power. How will we perceive when that power begins to reproduce harmful patterns?

But who will watch the AI?

Here our own proposal reaches its limit.

Paola Ricaurte, Edgar Gómez-Cruz, and Ignacio Siles warn that automated systems used by Latin American governments can reproduce inequality, coloniality, and authoritarian tendencies under an apparently neutral language of technological efficiency. (journals.sagepub.com)

Brazil's National Council of Justice already classifies systems capable of identifying profiles and behavioral patterns as high-risk applications in certain contexts and requires transparency, governance, non-discrimination, due process, and accountability in the judicial use of AI. (atos.cnj.jus.br)

Our Center would therefore need to be more auditable than many existing institutions.

Every query would leave a record.

Who accessed the system?

Why?

Which databases were used?

Which algorithm identified the association?

What evidence supported the alert?

Who authorized the next step?

And no AI could investigate, prosecute, and judge.

Brazilian scholar Mariana Mota Prado proposes thinking about anti-corruption systems through institutional redundancy: different organizations monitor, investigate, and decide, creating mechanisms similar to fail-safe systems. (doi.org)

Perhaps that is our answer.

Not to create the perfect watcher.

But to create a system in which:

no one can erase the memory alone,

no one can produce the accusation alone,

and

no one can decide the conviction alone.

Because independence without oversight can become power.

Oversight without independence can become persecution.

And intelligence without transparency can become surveillance.

In the Brazil 2026 Presidential Election, perhaps the question is not which of the three branches of government we want to strengthen.

Perhaps we can ask together:

How can we build a State capable of remembering for decades — including when patterns cross presidents, legislators, judges, companies, and investment funds - without allowing anyone to use that memory to decide alone who is guilty?

References

GARGARELLA, Roberto. O constitucionalismo brasileiro no contexto latino-americano. Suprema – Revista de Estudos Constitucionais, v. 4, n. 1, 2024. DOI 10.53798/suprema.2024.v4.n1.a413. (researchgate.net)

PRADO, Mariana Mota. Redundancy as a Legal Strategy to Combat Corruption: Exploring the Potential of Institutional Multiplicity to Create Fail-Safe Systems. Current Legal Problems, v. 77, 2024, pp. 335–376. DOI 10.1093/clp/cuae010. (doi.org)

PIEDRAHITA SIERRA, Sara. The Capture of Institutions as an International Phenomenon. Analecta Política, v. 13, n. 25, 2023. DOI 10.18566/apolit.v13n25.a08. (revistas.upb.edu.co)

RICAURTE, Paola; GÓMEZ-CRUZ, Edgar; SILES, Ignacio. Algorithmic governmentality in Latin America: Sociotechnical imaginaries, neocolonial soft power, and authoritarianism. Big Data & Society, 2024. DOI 10.1177/20539517241229697. (journals.sagepub.com)

BONFIM, Raul; LUZ, Joyce Hellen; VASQUEZ, Vitor. Mandatory Individual Amendments: a Change in the Pattern of Executive Dominance in the Brazilian Budgetary and Financial Cycle. Brazilian Political Science Review, 2023. (brazilianpoliticalsciencereview.org)

COAF — FINANCIAL ACTIVITIES CONTROL COUNCIL. Normative Instruction No. 1, July 16, 2026. Enhanced due diligence, ultimate beneficial ownership, corporate structures, and traceability. (gov.br)

NATIONAL COUNCIL OF JUSTICE. Resolution No. 615/2025, updated by Resolution No. 674/2026. Governance of artificial intelligence in the Brazilian Judiciary. (atos.cnj.jus.br)

FEDERAL COURT OF ACCOUNTS. TCU approves the President of the Republic's 2025 accounts with reservations. Brasília, June 10, 2026. (portal.tcu.gov.br)

OFFICE OF THE COMPTROLLER GENERAL. 2025–2027 Integrity and Anti-Corruption Plan. Transparency, integrity, expenditure oversight, and institutional strengthening. (gov.br)

SUPERIOR ELECTORAL COURT. Government Proposals — 2026 Elections. Official documents submitted by presidential candidates. (tse.jus.br)