Crime Does Not Begin in the Periphery - Primal Weichö, National AI, COAF, and the Capture of the State by the 0.01 percent
Crime Does Not Begin in the Periphery - Primal Weichö, National AI, COAF, and the Capture of the State by the 0.01 percent
Crime does not begin where it appears
Crime in Brazil does not begin in the periphery.
The periphery is often where crime appears most visibly: armed, territorialized, bloody, mediatic, and spectacular. It is where fear becomes routine. It is where bullets meet bodies. It is where police operations appear on television. It is where a poor young person becomes a suspect before becoming a citizen.
But this is only the visible surface of the machine.
Structural crime begins earlier. It begins when a tiny fraction of society is able to transform economic power into legislative influence, legal shielding, regulatory capture, money laundering, political financing, opaque lobbying, public contracts, investment funds, shell companies, millionaire legal defense, and the production of gray zones at the edge of the law.
The armed body in the periphery is often the last link in a chain.
The money, however, moves upward.
And when the money moves upward, it enters banks, fintechs, funds, real estate, fuel, betting markets, mining, foreign trade, campaigns, law firms, consulting firms, and institutional networks.
The central question is not only:
where is crime?
The central question is:
where does crime become respectability?
Acting on causes, not effects
The greatest mistake in public security is acting only on effects.
The effect is the armed young person.
The effect is the gang in the territory.
The effect is fear in the street.
The effect is the police operation.
The effect is the overcrowded prison.
The effect is the headline about violence.
But the cause comes before.
It is in the financial flow that launders money.
In the capture of the State by the few.
In the law written to measure.
In the absence of territorial return.
In the school that does not open inner worlds.
In childhood pressured by guilt, scarcity, and fear.
In religion used as control rather than liberation.
In politics that convinces people that humans are born evil in order to justify surveillance, punishment, and obedience from birth.
In Humankind: A Hopeful History, Rutger Bregman challenges the pessimistic view of human nature. A review by the London School of Economics describes the book as an attempt to dismantle classical assumptions that place humans as essentially selfish, arguing instead that our cooperative nature can guide better institutions. (blogs.lse.ac.uk)
This discussion is essential for public security.
Those who capture the State need to sell a negative anthropology.
They need to say that “man is evil by nature.”
They need to say that “we are born in sin.”
They need to say that bodily freedom is dangerous.
They need to say that youth must be tamed.
They need to say that the periphery must be contained.
They need to say that the poor must be watched.
They need to say that vulnerable territory must be occupied.
They need to say that living DNA must be controlled before it expresses itself.
This premise pressures the freedom of DNA from birth.
It makes the free expression of each axis, each inner creation, each possible world more difficult.
The child is not born only as risk.
The child is born as possibility.
Born as a sensitive body.
Born as a power of bonding.
Born as imagination.
Born as language in formation.
Born as an inner world opening.
Born as DNA wanting to experiment, belong, create, imitate, transform itself, and be recognized.
When society teaches too early that this body is guilty, dangerous, sinful, or suspicious, it begins to kidnap freedom before crime even appears.
This is the first invisible rifle: the narrative that points inward at the child and says that their nature must be defeated.
Primal Weichö: leaving the colonial local optimum
To act on causes, it is not enough to say that the human being is born good or evil. That opposition still belongs to the colonial local optimum. It traps the debate in the wrong question: should human nature be controlled or liberated?
In our concepts, we use primal Weichö to leave this local optimum.
The source from which we draw this concept is the exhibition Mundos Indígenas by Espaço do Conhecimento UFMG. The exhibition presents Indigenous worlds through concepts chosen by curators from five peoples: Yanomami, Ye’kwana, Xakriabá, Tikmũ’ũn/Maxakali, and Pataxoop. Among the concepts presented are në ropë, weichö, body-territory, yãy hã mĩy, and the great time of waters. The exhibition states that Indigenous worlds can teach us coexistence, respect, care, and the existence of other worlds. (ufmg.br)
In our extended view, primal Weichö is not only a word or cultural reference. It is a conceptual operator that liberates analysis from the colonial imprisonment that asks whether the human is born in sin, born evil, born dangerous, or born needing control.
That question has already kidnapped DNA.
It has already placed the living body under suspicion.
It has already transformed childhood into risk.
It has already made the free expression of each axis, each inner world, each possibility of creation more difficult.
Primal Weichö shifts the question.
It does not first ask whether the human is good or evil.
It asks:
what world is being allowed to be born in this body-territory?
It asks:
which relations are opening or closing the expression of living DNA?
It asks:
which colonial local optimum is trapping this body before it can create its own inner world?
It asks:
which systems of fear, guilt, sin, debt, scarcity, violence, and capture are preventing worlds from emerging?
In this sense, primal Weichö is the power before capture.
It is the field of opening where DNA can still create inner worlds before being reduced to sin, suspicion, productivity, consumption, obedience, or threat.
When religious movements claim that we are born in sin, they may produce a local optimum of guilt.
When political movements claim that the people are dangerous and must be contained, they may produce a local optimum of surveillance.
When the market claims that humans are fulfilled only through consumption, it may produce a local optimum of captured desire.
When the State claims that the periphery is the origin of crime, it may produce a local optimum of suspicion.
All these colonial local optima reduce primal Weichö.
They reduce the creation of inner worlds.
They reduce the freedom of living DNA.
They reduce the possibility of Real Jiwasa.
Therefore, acting on causes means protecting primal Weichö before it is captured.
The State sees the rifle, but is slow to see the financial flow
A State that moves from outside to inside does not reach the whole body-territory.
But it is also true that a State captured from above does not want to see the whole crime.
It sees the armed retail level.
It sees the drug point.
It sees the young man on the motorcycle.
It sees the favela.
It sees the border.
It sees the poor illegal miner.
It sees the vulnerable body.
But often it does not see, or takes too long to see, the financial flow that turns crime into economy.
It does not see the fund that launders.
It does not see the company that triangulates.
It does not see the fintech that hides.
It does not see the gas station that mixes fuel.
It does not see the importer that underinvoices.
It does not see the legal defense that eternalizes the process.
It does not see the law written to measure.
It does not see the lobby that creates exceptions.
It does not see the revolving door between public power, market, and private defense.
It does not see COAF data as a living system of causality.
Or, when it does see, the processes drag on until they lose social force, expire, fragment, or disappear from public memory.
This is the false Jiwasa of the captured State: society is summoned to fear visible crime, while structural crime gains time, technical language, expensive defense, influence, and respectability.
Organized crime is also organized market
When people speak of organized crime, the public image is usually that of armed territory. But crime is truly organized when it can cross borders between illegality and legality.
It needs accountants.
It needs lawyers.
It needs companies.
It needs invoices.
It needs fintechs.
It needs funds.
It needs fuel.
It needs real estate.
It needs contracts.
It needs technology.
It needs silence.
It needs influence.
It needs procedural time.
Recent operations in Brazil showed that criminal organizations may infiltrate formal sectors of the economy. In 2025, Brazilian authorities carried out operations against multibillion-dollar money laundering and fraud schemes linked to the fuel sector; Reuters described the action as a crackdown involving imports, domestic sales, fintechs, and investment funds. (reuters.com)
The Associated Press reported that Brazilian authorities seized about R$1.2 billion in assets in an investigation into laundering, corruption, and the fuel supply chain, involving suspicions around investment funds, shell companies, gas stations, ethanol plants, and payment institutions. (apnews.com)
These cases show that the simplistic distinction between “crime in the periphery” and “legal economy” no longer sustains serious analysis.
Crime is not only at the margins.
It tries to enter the center.
And when it enters the center, it needs technical, legal, financial, and institutional structure.
At that point, crime stops being only a police matter and becomes a matter of State architecture.
Fintechs, funds, and the new breathing of crime
Money is the breathing of crime.
Without clean money, crime remains limited. With laundered money, it buys real estate, companies, influence, logistics, lawyers, campaigns, protection, and reputation.
For this reason, the fight against crime cannot begin only with the rifle. It must begin with the financial flow.
Reuters reported in 2025 that Brazil’s Federal Revenue Service defended reviving reporting rules for fintechs amid signs that lesser-known payment institutions were being used for money laundering. The report also cited concerns about smuggling, cryptocurrencies, and online betting as instruments for financing organized crime. (reuters.com)
This shows a shift in scale.
Crime no longer depends only on hiding money in suitcases, straw men, or small businesses. It seeks digital systems, payment methods, corporate structures, funds, and financial technologies.
When financial infrastructure becomes faster than public oversight infrastructure, crime gains speed.
When the State does not integrate data, crime integrates.
When the State fragments competencies, crime connects operators.
When the State forgets processes, crime uses time as defense.
Lobby, legal capture, and crime at the edge of the law
Not everything that destroys body-territory appears as a legally defined crime. Often, structural violence is organized from within legality.
Economic sectors may pressure for laws, exceptions, amnesties, subsidies, tax rules, licenses, environmental flexibilization, pardons, temporal legal frameworks, special regimes, and regulatory loopholes. Part of this may occur legitimately in a democracy, if it is transparent and regulated. But when influence is opaque, concentrated, and disproportionate, it allows a few to write the field of play in which everyone else will have to live.
The OECD observed, in an integrity review of Brazil, that the country still lacks a specific framework defining lobbying and lobbying activities in a way that ensures transparency for these practices. (oecd.org)
Transparency International Brazil has also pointed to the need to regulate lobbying and strengthen anti-corruption, anti-money laundering, and oversight bodies, including COAF and the Federal Revenue Service. (transparenciainternacional.org.br)
The problem, then, is not only crime that violates the law.
It is also the structure of power that contracts the production of law in order to make violence legally manageable.
This is one of the deepest points of body-territory.
When law is made without body-territory, it can protect property against life, profit against the biome, contract against community, procedure against justice, and privilege against the future.
Millionaire legal defense and procedural asymmetry
Every person has the right to defense. Legal advocacy is not a crime. Technical defense is an essential part of the rule of law.
The problem is asymmetry.
When cases involving major economic interests can hire millionaire law firms, influential legal opinion writers, prestige networks, appeal specialists, delay strategies, and privileged access to the system, formal justice may continue to exist, but its material effectiveness becomes unequal.
Meanwhile, the poor body is judged quickly, held in pre-trial detention, exposed in the media, and treated as a threat.
Here a brutal difference appears:
poor crime is accelerated;
rich crime is sophisticated;
armed crime is televised;
financial crime is proceduralized;
street crime is urgent;
institutional crime is technical;
crime of the vulnerable body remains in public memory as danger;
crime at the top tries to expire in the process as complexity.
The problem is not the right to defense.
The problem is when millionaire defense combines with procedural time, influence, opacity, prestige relations, and institutional fragmentation to turn justice into a labyrinth.
The rule of law cannot be only the right of those who can pay to cross decades of litigation.
COAF: the body that should breathe with the State
If financial crime is the breathing of organized crime, COAF should be one of the intelligent breaths of the State.
COAF does not replace police, prosecutors, the Federal Revenue Service, the Central Bank, audit courts, or the judiciary. But it is a central piece for detecting suspicious operations, financial patterns, movement networks, and signs of money laundering.
For this reason, a country facing sophisticated crime cannot treat financial intelligence as secondary.
Transparency International Brazil published an analysis in 2026 stating that COAF’s budget had a real increase of only 0.56% between 2021 and 2024, which it considered modest given the size of Brazil’s financial system and the institution’s strategic relevance. (transparenciainternacional.org.br)
This reveals a civilizational contradiction.
Brazil demands maximum force against visible crime, but maintains limited capacity to confront the financial intelligence of invisible crime.
We want helicopters, armored vehicles, and spectacular operations.
But we do not give enough scale to the intelligence that follows money, links, companies, funds, fintechs, contracts, imports, real estate, fuel, betting, and influence networks.
Without a strong COAF, a strong Federal Revenue Service, an attentive Central Bank, an equipped Federal Police, technical prosecutors, and integrated courts, the State will continue wiping blood at the tip while money keeps breathing at the top.
National AI as a Fourth Power of procedural memory
AI has crystallized Logos.
Therefore, legal, financial, and institutional Logos cannot continue to be the monopoly of those who have money to eternalize proceedings.
A National AI, public, auditable, constitutionally limited, and subject to democratic control, could function as a fourth power of causal memory.
It would not be a robot judge.
It would not convict on its own.
It would not replace the defense, prosecutors, judges, courts, the Federal Revenue Service, COAF, or the Central Bank.
Its function would be different: to keep alive cases of high social, financial, environmental, political, and institutional impact.
Every case in which defense, consulting, lobbying, or legal representation cost above a certain threshold — for example, more than R$1 million — should enter a permanent regime of public memory and institutional intelligence, respecting legal secrecy, data protection, due process, and fundamental rights.
These cases could not simply disappear from the social radar.
They could be continuously analyzed, cross-referenced, and revisited with:
COAF reports;
public company and shareholder data;
public procurement data;
public contracts;
campaign donations;
environmental proceedings;
tax assessments;
import and export data;
corporate changes;
administrative sanctions;
judicial decisions;
public influence relations;
correlated legislative changes;
public social media of involved actors;
links among defendants, authorities, economic sectors, and law firms, always within legal, auditable, and republican parameters.
The goal would not be lynching.
It would be to prevent structural forgetting.
Today, sophisticated crime often wins through time. It wins through dispersion. It wins through technicality. It wins through data fragmentation. It wins because each agency sees only one piece. It wins because society forgets.
A National AI for procedural memory would permanently ask:
which major cases are sleeping?
which are approaching expiration?
which involve the same economic beneficiaries?
which law firms appear repeatedly in high-impact cases?
which legislative changes favored sectors under investigation?
which judicial decisions statistically deviate from comparable patterns?
which networks of influence appear in environmental, financial, tax, and criminal cases?
which body-territories were damaged while the case moved slowly?
This AI would not be the end of democracy.
It would be an attempt to prevent democracy from being captured by technical forgetting.
There are already signs that public AI is possible
The proposal for a National AI does not come from nowhere.
Brazil is already discussing uses of AI in public agencies and oversight systems. A 2024 paper on INACIA, an AI-assisted instruction system in Brazil’s Federal Court of Accounts, presented possibilities for using large language models to extract information from cases, analyze admissibility, and support recommendations in audit workflows, while recognizing limits and the need for caution. (arxiv.org)
There are also models and benchmarks focused on Brazilian legal Portuguese, such as Juru, trained with Brazilian legal sources, and LegalBench-BR, a public benchmark for evaluating models in Brazilian judicial decision classification. These works show that national and specialized legal AI is already technically plausible, although it must be used with public governance, auditing, and institutional limits. (arxiv.org)
This shows that the question is not whether AI will enter the State.
It is already entering.
The question is: will it enter to accelerate ordinary bureaucracy, or to confront the structural asymmetry that protects the top?
Will it enter to facilitate internal paperwork, or to rebuild institutional memory?
Will it enter as an efficiency tool, or as a public technology of body-territorial justice?
The National AI proposed here would have a higher mission: to prevent major cases from dying through social exhaustion, document dispersion, and technical privilege.
AI, money laundering, and complex networks
Money laundering is difficult to combat precisely because it operates in networks, across time, and under concealment.
A 2025 study on AMLGentex, aimed at research in anti-money laundering detection, highlights that laundering enables illicit funds to enter the legitimate economy, while only a small fraction is discovered. The study emphasizes challenges such as partial visibility for financial institutions, strategic behavior by launderers, rarity of confirmed cases, temporal dynamics, class imbalance, and network dependencies. (arxiv.org)
This is exactly why public and integrated AI is needed.
Financial crime is not a photograph.
It is a film.
It is not an isolated transaction.
It is a network.
It is not only an individual tax ID.
It is a constellation of individuals, companies, contracts, funds, straw persons, lawyers, operators, politicians, invoices, decisions, legal opinions, and flows.
An analog State faces digital networks at a disadvantage.
An auditable National AI could help see patterns that no isolated agency can see alone.
The capture of the State by the 0.01%
The thesis must be stated clearly:
crime in Brazil begins with the capture of the State by the 0.01% of the population when this minority is able to convert economic power into rules, shielding, procedural delay, tax exceptions, regulatory capture, and body-territorial destruction without return.
Here, “0.01%” is not only a literal statistic. It is a political image of extreme concentration: the top capable of transforming private interest into institutional design.
This does not mean that every elite person is criminal.
It does not mean that every businessperson, lawyer, judge, politician, investor, or public servant is captured.
The critique is not moral against individuals because of class.
The critique is structural.
When a few can influence rules disproportionately, escape the speed of ordinary justice, finance narratives, capture agencies, hire legal neutralization, and transfer costs to the people and the biome, the State ceases to be Real Jiwasa.
It becomes institutional false Jiwasa.
It mobilizes society to defend order, market, progress, security, and legality, but returns little to body-territory.
The 0.01% does not need to appear armed.
It appears as normality.
It appears as a legal opinion.
It appears as lobbying.
It appears as a legal thesis.
It appears as a tax exception.
It appears as consulting.
It appears as a fund.
It appears as influence.
It appears as delay.
It appears as expiration.
It appears as “complexity.”
And it is precisely this appearance of normality that makes structural crime so dangerous.
The armed young person is an effect; the captured State is a cause
The armed young person can be dangerous. They can kill. They can terrorize. They can dominate territory. They must be held accountable under the law.
But, in a body-territorial analysis, they must also be read as an effect.
An effect of a school that failed.
An effect of income that did not arrive.
An effect of a threatened family.
An effect of an abandoned territory.
An effect of a late State.
An effect of an illegal market that arrived first.
An effect of an economy that offers consumption without offering a future.
An effect of a country in which crime money moves upward into sophisticated structures while the vulnerable body remains exposed to bullets.
The armed young person is the scene.
The captured State is the structure.
Public security that targets only the scene can produce spectacle.
Public security that targets the structure needs COAF, the Federal Revenue Service, the Central Bank, National AI, audit courts, prosecutors, Federal Police, integrated data, transparency, whistleblower protection, lobbying control, procedural reform, and public memory.
Public security must begin with causes
Public security cannot begin in the periphery.
It must begin with financial flows.
With regulatory capture.
With money laundering.
With delay-based legal defense.
With fintech governance.
With opaque funds.
With public contracts.
With budget amendments.
With shell companies.
With fuel chains.
With illegal mining.
With land.
With real estate speculation.
With unregulated lobbying.
With procedural asymmetry.
With legalized environmental destruction.
But it must also begin before crime appears as violence.
It must begin with primal Weichö.
With childhood.
With the freedom of living DNA.
With the school that opens inner worlds.
With spirituality that liberates instead of blaming.
With politics that cares instead of frightening.
With culture that offers expression before the gang offers identity.
With territorial return before the illegal market offers money.
With protected Tekoha before the militia sells protection.
With Real Jiwasa before crime offers false belonging.
Then, yes, the State must enter armed territory when necessary, but not only with weapons.
It must enter with schools.
Health care.
Income.
Sanitation.
Community protection.
Intelligence.
Investigation.
Citizen DREX.
Clean energy.
Culture.
Public internet.
Local justice.
Productive credit.
Biome protection.
Material presence of the State.
The rifle without financial intelligence arrests the replaceable body.
National AI with procedural memory can reach the structure that makes the body replaceable.
Citizen DREX can arrive before false belonging.
Body-territorial education can open worlds before guilt, fear, consumption, or crime capture DNA.
COAF, National AI, Citizen DREX, and primal Weichö
COAF, National AI, Citizen DREX, and primal Weichö can form a new matrix of body-territorial State.
COAF detects suspicious flows.
National AI keeps memory alive, cross-references causalities, and prevents major cases from disappearing.
Citizen DREX returns material presence to body-territory before crime offers income, protection, or false belonging.
Primal Weichö protects the opening of inner worlds before the colonial local optimum reduces the human to sin, suspicion, productivity, or threat.
This matrix changes the question.
The question is no longer only:
how do we arrest more?
It becomes:
how do we prevent crime from buying the State, laundering money, and recruiting abandoned body-territory?
And, more deeply:
how do we protect the creation of inner worlds so that living DNA does not need to be born against the State, against the biome, or against itself?
This is true public security.
Not only punishing the effect.
But cutting the financial breathing of the cause.
And protecting living power before capture.
Metacognition for the reader
In the face of any news about crime, police operations, gangs, laundering, corruption, funds, fintechs, lobbying, proceedings, or judicial decisions, the reader can ask:
1. Is crime being shown only where it appears, or also where it is financed?
2. Does the report show the armed body, but also show the financial flow?
3. Who laundered the money?
4. Who provided legal defense?
5. Who legislated?
6. Who under-regulated?
7. Who delayed the process?
8. Who profited from delay?
9. Did COAF appear in the causal chain?
10. Does the damaged body-territory appear as victim or as suspect?
11. Does the proposed policy arrest only replaceable soldiers, or reach the architecture of crime?
12. Is AI being used to surveil the poor, or to make the top traceable?
13. Does the narrative begin from the premise that humans are born evil?
14. Which Weichö was blocked before crime appeared?
15. Which inner world failed to be born because the body was treated from the beginning as guilt, risk, or threat?
These questions are fundamental because they prevent society from being trained to hate only the visible tip.
Sophisticated crime wants the population to look downward.
Body-territorial metacognition forces the gaze upward.
And primal Weichö forces the gaze even deeper: before crime, before guilt, before suspicion, before capture.
Conclusion: crime appears in the periphery, but begins when the State is captured
Between crime and the rifle there is a lie: the idea that crime begins where it appears.
In reality, crime often appears in the periphery because it first passed through the center.
Through money.
Through contracts.
Through funds.
Through companies.
Through law.
Through legal opinions.
Through influence.
Through omission.
Through capture.
Through delay.
Through expiration.
Through shielding.
But there is an even earlier layer.
Before crime appears as a weapon, it may appear as a blocked world.
As childhood without opening.
As guilty DNA.
As a suspicious body.
As religion of control.
As politics of fear.
As school that does not create worlds.
As market that captures desire.
As a State that treats living power as risk.
For this reason, a body-territorial State cannot be only the State that arrives armed in vulnerable territory.
It must be the State that prevents the top from turning illegality into architecture.
It must be the State that uses public AI to keep the memory of major cases alive.
It must be the State that cross-references COAF, the Federal Revenue Service, corporate data, public contracts, influence networks, and damage to the biome.
It must be the State that knows that the armed young person is a visible effect, but the capture of the State by the 0.01% is a civilizational cause.
And it must be the State that protects primal Weichö: the possibility of each body-territory creating inner worlds before being captured by fear, guilt, scarcity, sin, market, or crime.
Visible crime asks for police.
Structural crime asks for the refounding of the State.
Ontological crime, the crime that kidnaps the power of DNA before life even flourishes, asks for a new civilization.
The New World only begins when body-territory stops being treated as suspect and becomes the measure of all justice.
Commented references after 2021
1. Espaço do Conhecimento UFMG. “Mundos Indígenas” - exhibition and catalog
Supports the conceptual source of Weichö, alongside në ropë, body-territory, yãy hã mĩy, and the great time of waters. The exhibition states that Indigenous worlds teach coexistence, respect, care, and the existence of other worlds. It is the source-base for the extended concept of primal Weichö. (ufmg.br)
2. Rutger Bregman. Humankind: A Hopeful History / LSE Review of Books - 2021
Supports the critique of negative anthropology: the idea that humans are naturally selfish, evil, or dangerous. Bregman’s reading helps dismantle symbolic foundations used by projects of control, punishment, and surveillance from birth. (blogs.lse.ac.uk)
3. Reuters. Operations against organized crime in Brazil’s fuel sector - 2025
Supports the thesis that Brazilian organized crime also operates in formal economic chains, such as fuel, imports, fintechs, funds, and laundering structures. It helps shift the analysis of crime from the periphery to the financial flow. (reuters.com)
4. Associated Press. Investigation into laundering, funds, and fuel - 2025
Supports the section on investment funds, shell companies, gas stations, ethanol plants, payment institutions, and seized assets. It shows how organized crime can infiltrate formal economic sectors. (apnews.com)
5. Reuters. Fintechs and money laundering concerns - 2025
Supports the need to treat fintechs, payment institutions, cryptocurrencies, betting, and smuggling as part of the possible new infrastructure for money laundering and organized crime financing. (reuters.com)
6. OECD. Integrity Review of Brazil - 2025
Supports the critique of Brazil’s lack of a specific and transparent framework for lobbying regulation. It helps ground the idea of legal capture and production of laws under unequal influence. (oecd.org)
7. Transparency International Brazil. Recommendations and unregulated lobbying - 2025
Supports the need to strengthen anti-corruption, anti-money laundering, and oversight bodies, including COAF and the Federal Revenue Service. It also reinforces the critique of disproportionate private influence in the political process. (transparenciainternacional.org.br)
8. Transparency International Brazil. “COAF sob pressão” - 2026
Supports COAF’s strategic importance and the critique of low budget capacity for financial intelligence in relation to the complexity of Brazil’s financial system. The analysis reports a real increase of only 0.56% in COAF’s budget between 2021 and 2024. (transparenciainternacional.org.br)
9. Jayr Pereira and collaborators. “INACIA: Integrating Large Language Models in Brazilian Audit Courts” - 2024
Supports the technical possibility of using large language models in public oversight agencies, especially Brazil’s Federal Court of Accounts, to support case analysis, information extraction, and recommendations, with caution and human validation. (arxiv.org)
10. Roseval Malaquias Junior, Ramon Pires, Roseli Romero, and Rodrigo Nogueira. “Juru: Legal Brazilian Large Language Model from Reputable Sources” - 2024
Supports the possibility of specialized Brazilian legal models trained on national legal sources. It reinforces the feasibility of a National Legal AI, provided it is auditable and institutionally limited. (arxiv.org)
11. Pedro Barbosa de Carvalho Neto. “LegalBench-BR” - 2026
Supports the need for Brazilian benchmarks to evaluate models in national legal tasks. It helps show that generalist models are not enough for the Brazilian legal system and that public AI needs local adaptation. (arxiv.org)
12. Johan Östman and collaborators. “AMLGentex: Mobilizing Data-Driven Research to Combat Money Laundering” - 2025
Supports the thesis that money laundering is a networked, temporal, strategic, and hard-to-detect phenomenon. The study helps defend public AI capable of cross-referencing data, identifying patterns, and keeping causalities alive. (arxiv.org)
