The STF, Artificial Intelligence and the Labor Court



The implementation of the replacement of the human being by AI presupposes that it has already been trained to act in the way desired by the ruling class


“Soon, I’m sure we will have artificial intelligence writing the first draft of sentences.”
(Luís Roberto Barroso)


It was with the above words that, in May this year, the President of the STF, Luís Roberto Barroso, expressed his project, disguised as a premonition, during J20, which is a meeting attended by presidents and representatives of the Supreme Courts of countries that are part of the G20 (

To justify his speech, the Minister resorted, once again, to the excessive judicialization that, according to him, exists in Brazil, adding that in the face of this problem, AI needs to be adopted to speed up the Justice system. In his words: “We have 85 million cases in Brazil at the moment, we have to have tools to speed things up.”

And he went further, stating that artificial intelligence “can make better decisions in many matters, because it is capable of processing more information at greater speed” (

In an apparent minimization of the effects of the idea, the Minister “recognized” that the use of technology involves risks and that, therefore, its application would depend on human supervision. “There are many risks and a discussion about regulating AI to protect fundamental rights, protect democracy” is necessary, especially because AI “still” cannot separate right from wrong or have common sense, added the Minister.

But was his concern really about preserving the role of human beings? What, after all, for the Minister, is right and wrong? What does he mean by “common sense”?

Amidst the challenge of analyzing this proposition regarding the use of AI in the Judiciary, I received, via social media, a video, where a “judge” boasted of having confirmed his hypothesis that a claimant, who was requesting recognition of the employment relationship in the period of five months before the actual annotation of his CTPS, he would have withheld the information in the initial petition that during that period he was receiving unemployment insurance and that, therefore, it had been his option to work without registration, to continue receiving the insurance -unemployment. When asked by the “interviewer” about the result he gave to the process, the judge, in a mocking tone, said that the lying complainant had committed an illegal act and that, therefore, the contract was null and void, with no right being granted to the worker. .

I was, of course, as much as indignant, but not at all surprised or disappointed by the response that was presented, and this was certainly the intention of whoever sent me the video. But, as I was already involved in this controversy about comparing humans to machines, I thought it opportune to ask ChatGPT the same question and I did so in the following terms:

“During the five-month period in which he was receiving unemployment insurance, a worker acquired a new job, but, because he was receiving the benefit, he was not registered. Subsequently, he filed a labor complaint requesting recognition of the contract for this period of 5 months and the receipt of labor rights corresponding to the period. The employer defended itself by saying that the law prohibits work during the period of receipt of unemployment insurance and that, therefore, the contract during that period would be null and void, meaning that the employment relationship and labor rights would not be recognized. Who is right?”

Here is ChatGPT's response:


“The situation presented involves complex questions about the legality of work while receiving unemployment insurance and the right to recognition of the employment relationship and corresponding labor rights. Let's analyze the arguments and applicable legislation.

Legislation on Unemployment Insurance and Employment Bond

1.                 Unemployment insurance:

· According to Law No. 7.998/1990, which regulates the Unemployment Insurance Program, the benefit is intended for unemployed workers who have no other source of income.

· Article 8, item III, of the aforementioned law establishes that the benefit will be canceled if it is found that the worker has another source of income.

2.                 Nullity of the Employment Contract:

· The employer argues that working while receiving unemployment insurance is illegal and therefore the employment contract would be void.

· However, Brazilian labor legislation, as per the understanding consolidated in several court decisions, generally does not consider the nullity of the employment contract as an argument to deny labor rights to the employee.

Jurisprudence and Principles of Labor Law

·                     Principle of the Primacy of Reality: In labor law, the reality of facts prevails over formal documents. If the worker actually provided services, the employment relationship must be recognized, regardless of the registration formality.

·                     Worker Protection: Labor legislation is protective in relation to workers, aiming to guarantee the basic rights arising from the employment relationship, even if there are formal irregularities.

Case Analysis

1.                 Recognition of the Employment Relationship:

· The worker effectively provided services during the five-month period, characterizing an employment relationship.

· Even if you received unemployment insurance improperly, this does not nullify the provision of services and, consequently, the employment relationship.

2.                 Corresponding Labor Rights:

· The worker has the right to recognition of the employment relationship and corresponding labor rights, such as salary, proportional vacations, proportional 13th salary, FGTS, among others.

3.                 Consequences for Unemployment Insurance:

· Improper receipt of unemployment insurance may result in the need to refund the amounts received to the government, and the worker may be subject to administrative sanctions.

· However, this does not affect the right to recognition of the employment relationship and labor rights arising from the employment relationship actually provided.


The worker is right to request recognition of the employment relationship and the labor rights corresponding to the five-month period in which he worked without registration. The effective provision of services characterizes the employment relationship, regardless of formal irregularities related to unemployment insurance.

Although the worker may have to refund the unemployment insurance amounts received unduly, this does not interfere with the right to recognition of the relationship and labor rights.

Legal References

·                     Law No. 7.998 / 1990: Regulates the Unemployment Insurance Program.

·                     Principle of the Primacy of Reality: Applicable in labor law, where the reality of facts prevails over formal documents.

Therefore, the employer must recognize the employment relationship and pay the labor rights owed to the worker, despite undue receipt of unemployment insurance.”



I confess that when I read the machine's response, I was a little shaken in relation to my preconceived conviction that I was completely opposed to AI. After all, the machine demonstrated much more refined legal knowledge and, in the specific context, reached a much more humanized result than those presented by the human being, an “expert” in Labor Law.

This forced me to go further with the experience, to even do a simulation regarding Minister Barroso's speech.

It seemed necessary, first of all, to ask the machine itself whether it would be capable of producing sentences.

The answer was as follows:


“Artificial intelligence (AI) drafting of sentences is an emerging and complex area within the legal field. The idea is that AI algorithms can help judges make judicial decisions, analyzing precedents, facts and applicable laws. However, this possibility comes with several challenges and ethical and practical considerations.

Possibilities of AI in Sentence Preparation

1.                 Analysis of Precedents and Legislation:

· AI algorithms can be trained to analyze large volumes of legal data, including court precedents and legal texts, to identify patterns and provide recommendations based on similar cases.

2.                 Bias Reduction:

· AI can help mitigate unconscious biases present in human decisions by basing its recommendations on objective data and established standards.

3.                 Operational:

· Automating part of the decision-making process can speed up the issuance of sentences, relieving judges' workload and reducing waiting time for the parties involved.

Practical examples

·                     Helsinki and Tallinn's AI Judge: In Estonia, a pilot project aims to create an “AI judge” to resolve small-value disputes (up to €7.000). The idea is that AI can issue decisions in simple cases, with the possibility of appeal to a human judge.

·                     COMPASS: In the US, the AI ​​system COMPAS (Correctional Offender Management Profiling for Alternative Sanctions) is used to help decide on parole. It analyzes data to predict the likelihood of a defendant reoffending, helping to decide on sentences and paroles.

Challenges and Ethical Considerations

1.                 Transparency and Explainability:

· One of the biggest challenges is ensuring that AI algorithms are transparent and that their decisions can be explained in an understandable way. The “black box” of algorithms can make it difficult to understand how the AI ​​arrived at a certain decision.

2.                 Inherent Biases:

· If AI training data contains historical biases, these biases can be replicated or even amplified in AI decisions. Therefore, it is crucial that the data used is carefully selected and audited.

3.                 Responsibility and Human Oversight:

· Judicial decisions have significant consequences for people's lives. Therefore, it is essential that the final responsibility for the decision remains with a human judge. AI should be seen as a support tool, not a complete replacement.

4.                 Data Protection:

· The use of AI implies the processing of large amounts of personal and sensitive data, requiring strict protection and privacy measures.


Although AI sentence crafting is technically possible and is already being explored in some contexts, there is still a long way to go for its widespread and safe adoption. AI can be a powerful tool for increasing efficiency and reducing bias in the judicial system, but it must be implemented carefully, ensuring human oversight and consideration of fundamental ethical principles.”



The machine, once again, surprised because it was much more cautious about its potential than the Supreme Minister, who referred to it as a panacea for all the problems of the Judiciary.

And it is important to realize that the machine did not present itself as a substitute for the human being, but rather as a tool that, through the use of algorithms and data collection, helps in the formulation of judicial decisions.

It is therefore appropriate to mention that there is a big difference between a system of algorithms and a system that reaches the level of artificial intelligence.

Although they start from the same principle, which is the storage and processing of data to reach a specific result, and are even referred to with the generic name of artificial intelligence, AI itself is a step further in relation to the algorithm system. The algorithms follow previously determined codes for collecting and “analyzing” data. AI is a system capable of producing its own codes, or, in a certain sense, of formulating an “autonomous rationality” and, through it, reaching the result relating to the demand proposed to it.

And then the questions expand indefinitely, because AI, obeying the purpose of providing an answer, to achieve its objective of interacting with human beings, does not exactly have a commitment to reality and, if necessary, creates its own path . The commitment, at least in this first stage of ChatGPT, is to formulate arguments that are similar to those of human beings, certainly using the information and data from which it feeds. It turns out that if the data is insufficient to reach the logical result, the AI ​​is capable of creating it, in order to maintain the argumentation structure typical of human beings. If these arguments require examples or conclusions based on factual and theoretical premises, ChatGPT, to satisfy the formal structuring, may create facts, precedents, authors and works.

To evaluate this performance, I asked AI questions about the topic of disconnection from work, through three different formulations, with the assumption that the answer had jurisprudential support, and in the three answers given, the processes indicated as jurisprudential paradigms simply did NOT existed, namely: AIRR-10800-79.2014.5.17.0013 TST; 0011359-98.2016.5.03.0180 TRT-3; 1000709-41.2018.5.02.0038 TRT2; and 1000123-89.2018.5.02.0007 TST.

This reminded me of how generative AI, linked to audiovisual, has been used, on social media, to create fictional characters or even to attribute speeches and gestures to real people that they have never expressed.

Specifically, AI has the power to create falsehoods (factual or theoretical) with the enormous appearance of reality.

This is a very serious problem, but it is not, and in fact will not, constitute an obstacle to the use of AI, especially because, no matter how many “merits” and “dangers” AI presents to us, it is just a a tool!

The fact is that the debate that takes place on the level of deification or demonization of AI is not at all rational and leads us to the false dilemma of the need to establish a position against or in favor of AI. After all, I repeat, AI is not a person, towards whom we direct feelings of affection or disgust. It's a machine.

And perhaps this is the main warning, because, given the enchantment provided by technological potential, we end up being induced – including through the influence of ideologically conceived media propaganda – to believe that technology explains itself, as if it were a phenomenon of nature. or that had a “life” of its own, and that produces results completely free of intentionality and/or partiality, therefore constructing insurmountable and indisputable determinations of human conduct.

Systems improve and invade people's daily lives, creating not only a dependency of the human being on the machine, but also an identity, which has increased considerably with the reproduction of machines in the – virtual – form of human beings and with the same bases of dialogue.

Thus, there is not only a process of submission to the inexorabilities brought by technology, but also, for example, the idea of ​​devaluing human work, as a kind of “humanization” of the machine. In many concrete situations in which the machine is placed in contrast to the human being, we find ourselves rooting for the triumph of the machine, as demonstrated, in an allegorical way, in the plot of the film “Ex machina: artificial instinct”, by director Alex Garland , from 2014. In the film, the director seeks to demonstrate that the evolution of AI can cause it to acquire the ability to deceive and develop a survival instinct. This, in fact, is not mere fiction, since, in concrete terms, artificial intelligence reproduces human “teachings” and values, and can therefore be trained to express, without effect and total control, both human virtues and weaknesses.

Within this increasingly intense context, we tend to attribute to machines the role of solving our problems and designing our future. When we do this, we are, in fact, reinforcing the romantic view that human beings are fallible by nature, after all, as they say, even with a certain pride, “to err is human”. To continue being human, we would have created the machine so that it could determine the paths to follow.

But if the machine is the concentration of human knowledge and experiences, it will tend, when processing this data, to also reproduce our “errors”, or even what we call “instinct” and it will do so, not through abstract rationality, superior, but based on the parameters that, in collective training, appear to her as the dominant one.

Thus, we reach the inevitable point of stating that whoever dominates technology – because it, in the capitalist production model, based on property rights, has an owner – will be fully capable of developing training mechanisms for the machine so that it reproduces, as correctly , your own worldview.

But, it is worth insisting, this perception will not be enough, as long as this model of society persists, to prevent AI from being introduced into our reality. Hence why, understanding that this system only works based on the multiplicity of data and information that is introduced to it, which requires an unrestricted publicity of access and interactions, the task of control is not so simple and, consequently, the universe of AI presents itself as yet another field of dispute.

When one points to the character of domination that AI possesses and simply abhors the idea, the only effect that is produced is the facilitation of the path of yet another mechanism of oppression, which is even capable, given its “enchantment”, to legitimize and naturalize discrimination, prejudice, exploitation, exclusion and inequalities.

It is certainly for this reason, in other words, the fact that it has not yet been completely trained, or, in other words, tamed, that Minister Barroso says that the AI ​​is not yet capable of producing sentences. And what it advocates – with the generic name of AI –, in reality, is merely the use of a system of algorithms so that the STF's understanding is necessarily adopted by the Judiciary as a whole.

Minister Barroso's proposal, specifically, is linked to the commitments made by the Brazilian Judiciary since the implementation of Technical Document no. 319, from the World Bank, from 1996, which had the specific topic: “The Judiciary Sector in Latin America and the Caribbean – Elements for Reform”, prepared by Maria Dakolias, appointed as “specialist in the Judiciary Sector of the Private and Public Sector Division of Modernization”.

In this document, the conclusion reached by the “experts” is that there is a pressing need to “rethink the role of the State”. Textually, the document refers to “greater confidence in the market and the private sector, with the state acting as an important facilitator and regulator of private sector development activities”. This fact determines, according to the document's signatories, a “need for reforms to improve the quality and efficiency of Justice, fostering an environment conducive to trade, financing and investments”. In this context, in which the State must appear as a mere facilitator of the economy, a Judiciary is proposed that interprets and applies laws in a “predictable and efficient” way, and that values ​​the composition of disputes.

Let's not let ourselves be fooled, then. The Minister's speech is, rather, a nod to the market, reinforcing the speeches of the minimum State and jurisdictional control to provide a certain “legal security” guided by precepts that are capable of guaranteeing greater profitability for foreign investments in the country. It therefore represents a minimization of the relevance of public servants and an affront to the powers and independence of judges.  

It is worth noting that the argument in question was used precisely at a time when an administrative reform aimed at reducing the role of the State was demanded and public servants presented themselves as a relevant force of resistance.

It is also necessary to verify that, ideally, the improvement of work tools should be at the service of reducing the burden of work and reducing the hours of life dedicated to work, but, in the Judiciary, electronic tools have been used to increase the workload, especially due to the greater possibilities of “online” control over the achievement of goals that are established in parameters that are known to be unattainable, creating a sickening work environment that destroys self-esteem.

The speed advocated, linked to the market demand for stability and predictability of business, does not envisage the realization of rights, but the elimination of legal uncertainties, based on a unique standard of understanding the law committed to the pains of capital, especially because, in In an economic reality marked by precariousness, speed in itself does not represent a concrete element in the implementation of judicial sentences, which remains an unattainable step for the majority of creditors. (about unsatisfied executions, see the text:

There is, therefore, no concern with improving judicial provision and the effectiveness of rights, especially social rights. A true concern in this sense would begin by asking what are the real reasons why there is a large legal demand, which even presupposes questioning, given the possible hypotheses, whether, in fact, there are actually many demands or, on the contrary, if what we experience is a situation of repressed demands and lack of access to Fundamental Social Rights.

And it is extremely important to highlight that this topic brings to light relevant issues linked to national sovereignty itself.

To fulfill the commitment signed by Brazil with the World Bank, the latter has invested enormous amounts in structuring management and virtualization mechanisms in the Brazilian Judiciary. In 2012, for example, the CNJ announced the conclusion of an agreement with the World Bank, through which the latter donated the amount of 450 thousand dollars to the Brazilian Judiciary, for “carrying out a study that points out the asymmetries that exist in Brazilian state justice , in terms of management, training and computerization” ( According to the CNJ, the implementation of the computerized system in the Courts would be carried out free of charge (, but there are many reservations about this, as can be seen from the news (;

Furthermore, at least in terms of information contained on the internet, nothing is very clear in terms of the possible transfer, to companies that own the technologies, including the virtualization of hearings and sessions, of the personal data of litigants and their lawyers( as), in addition to the content of the conflicts.

The issue, as we can see, is very far from being just a simple discussion around the scope and way of applying AI to carry out judicial acts. It concerns, precisely, the increasingly urgent dilemmas linked to the dispute surrounding the production of consciousness, knowledge and values ​​constitutive of the human condition.

Who will AI tools serve and what purposes will they serve?

The answers, I insist, depend greatly on our ability to intervene in this dispute.

In a more concrete way, remembering the obstacles that, through human action, have repeatedly been placed in the implementation of Social Rights and, more notably, Labor Rights, in our reality, we can start by asking: what parameter of human being have we conceived?

Facing the challenge of seeking an answer, it is important to remember that human rationality, from which Social Rights resulted, was produced at the end of two world wars, when the successive errors committed throughout the 1919th century were recognized. From this assessment, expressed in the numerous documents that preceded the creation of the ILO (in XNUMX), a consensus emerged around the implementation of limits to the classical liberal precepts of individualism and free competition. Social Rights and Labor Rights, in particular, are the explanation of the understanding surrounding the need to establish such restraints and the ways in which these are implemented were enshrined in precise formulas linked, for example, to the limitation of working hours, minimum age for work, salary sufficient to guarantee the dignified existence of the worker and his/her family, prohibition of any type of discrimination, rest periods, guarantee of survival in case of unemployment, etc.

On that occasion, commitments were also made regarding the effective international applicability of the norms then conceived and which, for the purpose of full effectiveness, were integrated into a system marked by some fundamental precepts, such as: the public order nature of the norms; the minimum nature of the guarantees set; the non-renounceability of rights on the part of their holders; and the rule of the primacy of reality, in the assessment of the legal effects of a given form of labor exploitation.

However, what we saw – with greater intensity in recent years – were human beings creating arguments to break the pact, but always with the production of a rationality aimed at disguising intentionality. The reason used to pervert the real and not for it to be understood and analyzed.

Even with different norms with express and unequivocal statements in a certain direction, a completely opposite and quite minimized approach was reached, almost always appealing to “reasonableness” and “common sense”.

In Labor Law, the argumentative strategy was to accuse the rigidity of the rules, which were even created for an outdated reality. It was up to the interpreter and applicator of the standards to give them meanings more adapted to the demands of the “modern” world, establishing the assumption of a necessary “flexibility” as an authentic principle. In Brazil, with the objective of weakening labor achievements, since the 50s – with deepening in the years of the civil-business-military dictatorship, the neoliberal exacerbation (90s), the democratic rupture in the Temer period and ultra-liberal politics and denialist of the Bolsonaro government –, the rhetorical violence was spread that labor laws, in addition to generating high costs for production, even hindering the “generation of jobs”, were created under the fascist Vargas regime and constitute an obstacle to the economic development and even the freedom of workers. The costs of labor rights in Brazil, which are unparalleled in other countries, would be the cause of unemployment and “informality”

At the time I was writing the paragraph above, I received a new message online, in which the editorial of the newspaper O Globo published today (17/05/24) was reproduced. The content of the editorial is such an exact copy of this old primer (using identical – and already refuted countless times – arguments and seeking support from the same academic “authority”) that it is worth reproducing it, with the warning that it does not It was produced by Artificial Intelligence, which is, rather, the explanation of the human capacity to deceive and formulate logic based on invented facts:



“Labor legislation continues to weigh against job creation – There is an unequivocal relationship between the high cost of creating jobs with a formal contract and high informality

While the Executive, Legislative and Judiciary discuss how to tax company payrolls, few remember the main fact surrounding the issue: employment in Brazil is expensive. The employer, in addition to paying its employee, needs to spend the equivalent of a second salary on contributions to Social Security, the Guarantee Fund for Length of Service, education salary, thirteenth salary, vacations, accident insurance, contributions to System S, etc.

According to José Pastore, a researcher at the Fundação Instituto de Pesquisas Econômicas (Fipe), an employee with a formal contract costs the employer, adding up all the resulting rights, 103,7% of the salary. An industry, when hiring a worker at the average salary paid by the sector to those who have completed secondary education, R$2.287, will have to spend another R$2.371,62 on fees. In his words, workers “earn little and cost a lot”.

This is one of the largest proportions in the world. Considering only taxes on salaries and Social Security contributions — excluding charges such as vacations, thirteenth salary and other taxes —, Brazil is second only to France in a list of 42 countries, according to the Organization for Economic Cooperation and Development (OECD). At 25,8%, it exceeds the OECD average (13,8%), Germany (16,5%), Mexico (10,4%), United Kingdom (9,8%), China (22,1%) and the United States (7,6%).

It is no coincidence that the United States has a robust job market. The relationship between labor costs and informality is unequivocal, as charges paid to the government act as a disincentive to job creation. There is no other reason for there to be so much informal work in Brazil. Those who do not have a formal contract — nor access to benefits such as vacation or thirteenth salary — represent 38% of the workforce, or 38,8 million, according to IBGE.

The Consolidation of Labor Laws (CLT) was enacted by Decree-Law by Getúlio Vargas in 1943, still during the Estado Novo dictatorship. It included labor rights of all types, created at a time when Brazil was still more rural than urban. The labor reform promoted in 2017 during the Temer government was successful in making several aspects of this archaic legislation more flexible. Even so, Brazilian law still imposes obstacles to the generation of jobs and wealth. They need to be removed.

It is not a question, as union leaders argue, of “making jobs precarious”, but of adapting them to the conditions of a modern economy. The precariousness arises from the high weight of contributions that the employer has to bear when creating jobs with a formal contract. Reducing it will strengthen the job market and provide greater economic growth.”



In 1999, when I completed my undergraduate thesis, “Labor Law as an Instrument of Social Justice”, I pointed out that the idea of ​​flexibility, fueled by the inexorabilities of “globalization”, was just a rhetorical argument to facilitate the objective of disregarding the literalness and logical coherence of labor standards. Therefore, the “revolutionary” role of the labor jurist concerned with the issues, dilemmas and anxieties of the working class verifiable in the context of the sale of labor power for survival, would be to merely apply the norms in their strict literalness and with respect minimum to systemic coherence. This is because the normative set of a social nature already represented, as stated, the result of an understanding, historically conceived, around the existential minimum that should be granted to male and female workers.

There was, and still isn't, anything to argue about this, therefore.

However, as warned in the same text, the mostly bourgeois worldview of interpreters and enforcers of labor law negatively interfered with the effectiveness of workers' rights.

And, in another text, I tried to demonstrate that labor legislation was a heritage of the working class, which could not be usurped by non-working class law enforcers.

What I advocated, if we think about it carefully, looking at the issue we are dealing with, was an “automated” application of legislation, thus preventing human action, influenced by the bourgeois worldview, from being an instrument to divert labor standards of its objectivity.

Updating the debate, we would have the plan of human action marked, as defended by Minister Barroso, by common sense, in which the correct answer to a legal question would be produced, and the universe of automated AI intervention, which would only be considered after being validated by human beings.

To better evaluate this confrontation and even test my thesis about how the worldview linked to the dominant class of interpreters and applicators of Law constitutes an obstacle to the implementation of Social Rights, I thought about the current issue regarding the competence of the Labor Court, which, As everyone knows, it has been emptied by repeated decisions of the STF.

But what would AI say about this?

First, I asked ChatGPT if it is possible, in the context of a Constitutional Complaint, for the STF to reform the Court's decision that involves the assessment of factual matters.

The answer was as follows:



“In the context of the Brazilian legal system, the Constitutional Complaint is an instrument designed to preserve the jurisdiction of the Federal Supreme Court (STF) and guarantee the authority of its decisions. However, regarding the possibility of the STF reforming Court decisions that involve assessment of factual matters, the situation is more complex.

As a rule, the STF is not a review body for facts, but rather for questions of law. This means that, in principle, the STF does not re-evaluate the factual matter already decided by the lower courts, focusing only on the application and interpretation of the law. This guidance is based on STF Summary 279, which states: “For a simple re-examination of evidence, there is no extraordinary appeal.”

However, there are exceptions to this rule. In situations where the contested decision directly contradicts the decision-making authority of the STF or where there is disrespect for a binding precedent, the Constitutional Complaint may be upheld, even if the analysis indirectly involves a review of factual matters. The objective here is not to reevaluate the facts themselves, but to ensure uniformity of constitutional interpretation.

For example, if a lower court made a decision based on an interpretation that conflicts with a STF precedent or a binding precedent, the STF may, when ruling on the Complaint, reverse that decision to maintain the integrity of its jurisprudential understanding. In this context, the assessment of factual matters may occur in an ancillary manner, only to the extent necessary to ensure the correct application of the constitutional norm or consolidated jurisprudence.

Therefore, although the review of factual matters is not the main focus of the Constitutional Complaint, the STF may, in exceptional cases, review decisions involving factual matters if this is essential to guarantee the authority of its decisions and the correct application of constitutional law.

In short, the reform of Court decisions by the STF, in the context of a Constitutional Complaint, is not common when it involves pure factual matters. However, it can occur in situations where the factual matter is intrinsically linked to disrespect for a decision of the STF or a binding summary, always with the objective of preserving the integrity of the legal system and the authority of the STF.”



Well, as is well known, the STF does not have any decision regarding jurisdictional jurisdiction when debating whether there is an employment relationship in a given employment relationship and, therefore, this debate in itself would not challenge the STF's authority. There is not even a decision by the STF stating that hiring for self-employment excludes this assessment and automatically shifts the jurisdiction to the Common Justice, except in the monocratic decisions made in the Constitutional Complaints and, obviously, a decision that deviates from the constitutional parameter it cannot serve as a precedent to legitimize other deviations.

The STF's only decision, always referred to in the Complaints, concerns the validity of the outsourcing.

So, I asked the machine if there was, for example, a difference between hiring by a PJ and outsourcing.

The machine said:


“Yes, there are significant differences between hiring a legal entity (PJ) and outsourcing, although both arrangements may involve the provision of services by companies or professionals not directly employed by the contractor. Let's explore each of these concepts in detail:

Hiring a Legal Entity (PJ)


· Hiring a PJ occurs when a company or entity hires another company (which may consist of a single professional or several) to provide certain services.


·                     Runtime: The contracted PJ generally has greater autonomy in the execution of services, being able to define how and when it will carry out the activities, as long as it respects the deadlines and conditions established in the contract.

·                     Focus on Delivery: The contract is usually based on the delivery of specific results or the completion of defined projects.

·                     Nature of the Contract: The relationship is governed by a commercial contract, without an employment relationship.

·                     Responsibilities: The hired PJ is responsible for its own labor, social security and tax costs.

·                     Risks and Benefits: The PJ assumes business risks and can obtain tax and operational benefits depending on its structure and tax regime.



· Outsourcing is the hiring of a specialized company to carry out secondary activities (activities not essential to the contractor's core business) or even core activities, as permitted by law, through its employees.


·                     Subordination: Outsourced workers are employees of the outsourcing company, but work under the supervision and control of the contractor, especially in relation to the activities to be performed.

·                     Intermediation: The outsourcing company mediates the employment relationship, being responsible for hiring, paying salaries, benefits and fulfilling labor and social security obligations.

·                     Continuous Services: Outsourcing usually involves the provision of ongoing services of an operational nature, such as cleaning, security, maintenance, etc.

·                     Nature of the Contract: The contract is between the two companies (contractor and outsourcer) and focuses on the services provided, but the workers involved have an employment relationship with the outsourcer.

Main Differences

1.                 Employment Relationship:

·                    PJ hiring: There is no employment relationship between the contractor and the contracted PJ professionals.

·                    Outsourcing: The workers have an employment relationship with the outsourcing company.

2.                 Runtime:

·                    PJ hiring: The contracted PJ has greater autonomy in executing the services.

·                    Outsourcing: Outsourced workers work under the direct supervision of the contractor.

3.                 Labor and Social Security Responsibility:

·                     PJ hiring: The hired PJ is responsible for their labor and social security expenses.

·                      Outsourcing: The outsourcing company is responsible for the labor and social security costs of its employees.

4.                 Kind of service:

·                     PJ hiring: It usually involves specific projects or delivery of determined results.

·                     Outsourcing: Often related to ongoing services and operational activities.

Legal Considerations

Brazilian legislation, especially after the 2017 Labor Reform (Law 13.467/2017), introduced specific regulations for both the hiring of PJs and outsourcing, aiming to reduce the precariousness of work and guarantee the rights of outsourced workers.

In summary, although both arrangements can be ways for a company to obtain external services, corporate hiring and outsourcing differ significantly in terms of employment relationship, autonomy, responsibilities and types of services provided.”



As expressed by ChatGPT, in outsourcing the employment relationship between the worker and the outsourced work company is presupposed, and outsourcing concerns, more specifically, the link between the providing and receiving companies, notably, regarding the responsibility of the latter vis-à-vis the employees of the former. Hiring by a PJ, in which the worker himself is the “company” providing the service, with separation from the employment relationship, is in no way to be confused with outsourcing, and therefore, when examining the case involving possible fraud in hiring by a PJ, the STF precedent that pointed to the validity of outsourcing.

The STF, therefore, according to the machine, is using the Constitutional Complaint improperly, as it has been reforming decisions that involve factual matters, without the assumption of preserving the decision-making authority of the STF itself.

But I went ahead and asked: does hiring a worker as a self-employed worker prevent the possible existence of an employment relationship, that is, subordinate work, in the provision of services carried out?

The answer:


“The hiring of a worker as an independent contractor does not prevent, specifically, the possible existence of an employment relationship from being verified. In Brazilian labor law, the employment relationship is defined based on substantive criteria and not just the nomenclature or form of the contract. These criteria are established in the Consolidation of Labor Laws (CLT) and consist of the following elements:

1.                 Personality: The work must be performed personally by the worker and cannot be transferred to another person.

2.                 Habituality: Work must be provided continuously, and not sporadically or occasionally.

3.                 Onerosity: The worker must receive remuneration for the work performed.

4.                 Subordination: The worker must be under the command and direction of the employer, receiving orders as to how, when and where work is to be carried out.

Even if a worker is hired as an independent contractor, if, in practice, the employment relationship presents these elements, it can be characterized as an employment relationship. The name given to the contract or the legal form used for contracting is not decisive. What matters is the reality of service provision.

Labor Relationship Analysis

To assess whether an employment relationship exists, the Labor Court analyzes the specific facts and circumstances of the case. Here are some points to consider:

·                     Control and Direction: If the worker is subject to rigid schedules, detailed rules on how to perform their tasks and constant supervision, this may indicate subordination.

·                     Integration into the Main Activity: If the worker performs activities essential to the company's main business, this may indicate that he or she is not truly self-employed.

·                     Exclusivity: If the worker provides services exclusively for a contractor, this may be an indication of an employment relationship, although it is not conclusive in itself.

·                     Economic Dependence: If the worker is economically dependent on the contractor, this may be an indication of an employment relationship.

Jurisprudence and Judicial Practice

The Labor Court, in several decisions, has recognized the employment relationship even when the contract was formalized as a self-employed worker, if the factual elements demonstrate the presence of the requirements of the employment relationship. The analysis is always case-by-case and considers the peculiarities of each case.

Consequences of Requalification

If the existence of an employment relationship is found to exist, several legal consequences may occur:

·                     Registration at CTPS: The employer will be obliged to register the worker as an employee, with a note on the Work and Social Security Card (CTPS).

·                     Retroactive Payments: The employer may be ordered to pay all due labor payments, such as vacation, 13th salary, FGTS, overtime, among others.

·                     Fines and Penalties: The employer may suffer administrative fines and be held liable for violations of labor standards.

In short, hiring a worker as a freelancer does not, in itself, prevent the verification of an employment relationship. The reality of the provision of services is what defines the nature of the relationship. If the elements of the employment relationship are present, the Labor Court can requalify the relationship, recognizing the corresponding labor rights.”



At this moment, with all the pieces in place, the opportunity came for a final move.

I confess that I was a little fearful, but I still asked: in Brazil, which segment of the Judiciary is competent to say whether there is an employment relationship in a given employment relationship or not?

The answer given by the machine was unequivocal:


“In Brazil, the segment of the Judiciary competent to say whether, in a given employment relationship, there is or is not an employment relationship is the work justice. The Labor Court is responsible for judging disputes between employees and employers and other controversies arising from labor relations, as established by the 1988 Federal Constitution.

Structure of the Labor Court

The Labor Court is made up of three levels:

1.                 Labor Courts: They are the first instance and receive labor complaints. In these Courts, labor judges initially judge cases, including verifying the existence or not of an employment relationship.

2.                 Regional Labor Courts (TRTs): They are the second instance and have jurisdiction over one or more units of the federation. The TRTs judge appeals against the decisions of the Labor Courts and can also judge other labor matters within their original jurisdiction.

3.                 Superior Labor Court (TST): It is the highest instance of the Labor Court. It judges appeals against TRT decisions and has the function of standardizing labor jurisprudence across the country.

Jurisdiction of the Labor Court

The competence of the Labor Court to recognize the existence or not of an employment relationship is provided for in article 114 of the 1988 Federal Constitution, which establishes:

“The Labor Court is responsible for processing and judging: I – actions arising from the employment relationship, covering entities governed by external public law and direct and indirect public administration of the Union, the States, the Federal District and the Municipalities.”

Furthermore, the Consolidation of Labor Laws (CLT) and various infra-constitutional norms also regulate the duties and procedures of the Labor Court.

Employment Relationship Recognition Process

When a worker files a labor complaint alleging the existence of an employment relationship that was not recognized by the employer, it is up to the labor judge at the competent Labor Court to analyze the evidence and arguments presented by the parties. This judge will assess whether the requirements of the employment relationship — personality, habituality, onerousness and subordination — are present in the specific case.

If the judge recognizes the existence of the employment relationship, he will determine the legal consequences, such as registration in the Work and Social Security Card (CTPS) and payment of the labor funds owed.

Resource and Review

The decisions of the Labor Courts may be subject to ordinary appeal to the Regional Labor Courts, and, subsequently, to the Superior Labor Court, if there are relevant questions of law that justify the intervention of the higher court.

In summary, the work justice is the competent segment of the Brazilian Judiciary to determine whether or not there is an employment relationship in a given work relationship, and it is up to it to judge and decide on conflicts arising from these relationships.”




The machine won and someone more hasty, even accepting the suggestion made by Barroso himself, would already propose replacing the STF Ministers with Artificial Intelligence. At least in labor matters, for now, the decisions made by the machine would be much more in line with the legal order than those that have been pronounced within the scope of the STF.

However, despite having seen confirmed the thesis that without the interference of “reasonableness” and “common sense”, that is, applying the legislation in its strict literality, with systemic coherence and in accordance with its principled assumptions, the result that is produced, going in the direction of greater protection of male and female workers, limiting the interests of employers, is one that is logically extracted from the current normative set, but this result, it must be recognized, represents a ephemeral victory, since, faced with the answers presented by the machine, the detractors of labor legislation who hold the power of the last word, will soon try to train the algorithms so that such answers are no longer given, at least within the scope of the Judiciary .

It is true that the experiment carried out above is still an important demonstration of how AI tools can be used to direct the ideological dispute and, above all, to reveal what, in effect, would be the intentions that are hidden behind the defense of the inexorability of technological evolution.

On the other hand, when we feel contemplated by the answers given by the machine, we need to recognize that this, in fact, explains how much we are reducing our own trust in human knowledge, transferring argumentative authority to the machine.

This phenomenon is not just the result of a certain mental laziness, as machines have done much of what would be the result of human mental effort. But it is also the result of the perception that the knowledge production process itself has been commodified, and is therefore, in many situations, under the control of economic power.

Certainly, many people, reading this text, felt their “souls washed” when they saw that the machine expressed the answer they wanted to hear/read. It turns out that so many of us had already reached this same result, with bigger and better arguments. However, the human argument does not reach the necessary potential for widespread conviction, since we are under the logic of knowledge dominated by ideological disguises and we transfer the attribute of impartiality to the machine.

We need to reflect on this, as it reveals a certain assimilation of human failure.

This is why it is a huge problem to let ourselves be invaded by the momentary satisfaction of being able to say that the machine has proven to be much more intelligent and much less biased than certain human beings, and, with this feeling, come out in defense of the use of AI in the Judiciary, especially because this use, in terms of mere argumentation, is quite different from another that can be established institutionally, developed based on parameters controlled by a top group of administrators.

This forces us to return to the essential question surrounding the challenges of constructing values ​​for the constitution of the human condition, because only from human conceptions linked to the recognition of effective, real and full equality between human beings will we be able to understand, including the processes of manipulating the machine and opposing it and its eventual dominators, with arguments that reflect an effective concern for humanity.

The challenge is quite big because in this model of society, in which the appearance of relationships leads and even shapes our feelings, as we are hallucinated by the fetish of the goods we produce and which we consume often even without it being to satisfy any concrete need, the worldview that tends towards generalization is the one that satisfies the interests of the rule of capital. This means that the understandings of the world that are relevant to the project of overcoming class society and eliminating the various forms of discrimination and oppression are subjected to a process of destruction and disqualification, through disguises and argumentative dissimulations, which, even for If they pretend to be objective, impartial, scientific and, therefore, indisputable, they currently use the authority of AI.

This is why the dispute over conscience, necessary to the class struggle, cannot leave the field of virtuality, to prevent false “truths” from becoming naturalized at the service of human beings who, at all costs, just want to promote themselves, earn social privileges and economic advantages and satisfy feelings of vanity, which presupposes obtaining recognition from dominant powers, maintaining social inequalities and destroying all forms of resistance and social advancement, not being constrained even by environmental limitations.

Therefore, the experiments above say more about the devices concretely used by certain human beings to achieve and satisfy their unrevealed goals and interests than about the merits of the machine.

It is, even less, a possible conflict between human beings and artificial intelligence.

The machine, after all, only expressed a result that had not yet been manipulated, extracted from the literal rules applicable to the cases about which it was questioned, positioned in a logical and coherent way.

The conclusion reached is that the task of effectively human beings around the construction of a society effectively compatible with the human condition is in full force and with increasing challenges, with virtual confrontation being, perhaps, the main one. . If at the time of Marx, proletarian mobilization in the streets, arising from a consciousness produced from the dissemination of critical knowledge expressed in pamphlets and doctrinal works, was the only form of struggle, today, in addition to this, virtual dispute is necessary , seeking to avoid virtual manipulation and, with this same weapon, attack artificialized concrete.

The fact is that we will never be able to transfer the challenges and tasks imposed on us to the machine, which begins with learning ancient knowledge, notably from marginalized and almost decimated cultures, to produce knowledge that is effectively committed to overcoming inequalities. and all forms of oppression, prejudice and discrimination of a racial, gender and ableist nature.

In a reality dominated by the virtualization of life, where sharing images, obtaining “likes”, spreading hate and satisfaction with “cancellations” drive actions, also marked by extreme individualism, our responsibility to make these confrontations linked to the essence of a model of society in which human beings were objectified and fetishized, developing relationships only on a superficial level, all in the service of maintaining the dominance of a few over many others.

Our horizon, for the construction of the human condition, more than ever, must be to overcome this model of society.

And the machine can be both an instrument of this fight and a powerful weapon for the opponent!

Along this path, we must, with all our strength, fight against the replacement of the human being by AI, since when this happens, maintaining the same social, political and economic arrangements structuring the production and consumption of goods, for the reproduction of capital, therefore, substitution functions only as a mechanism for devaluing work and women, in order to maintain and even deepen inequalities, including reinforcing gender and racial oppression and the consecration of individualism and of ableism, we are not facing a mere and inevitable effect of technological evolution but rather the triumph of dominant forces.

The implementation of the replacement of human beings by AI presupposes that it has already been trained to act in the way desired by the ruling class.

And it cannot be forgotten that in the international organization of productive forces, Brazil is in a position of dependence, meaning that the empire of an AI configured along the lines of the interests of foreign companies that maintain a monopoly on this knowledge also represents a form to reaffirm colonialism.

It is evident that the effects of this domination strategy can get out of control and go well beyond what was predicted and desired by those who implement it. But before any positive effects can be glimpsed from this detachment, we will have to experience the disastrous results of human bankruptcy and economic and social chaos, with environmental destruction, naturalization of violence and multiplication of suffering, hunger and conflicts of all kinds.

Something very similar to what we are already experiencing, or not?

The truly intelligent thing, therefore, is to improve artificial intelligence so that it is at the service of human beings, in order to eliminate work as a workforce from their existence, thus giving them back humanity.

The first and decisive step in this direction, considering the issue within the strict limits of the proposition addressed in this text, is to extract AI from the domain of private companies, which, through representatives that we need to know how to identify, use the power derived from this knowledge to make money and disseminate, reinforce and feed the various forms of exploitation and oppression.

And in which socio-political-economic arrangement will it be possible to reach the full stage of using AI in the service of humanity?

The answer, my friends, is once again blowing in the wind!

*Jorge Luiz Souto Maior is a professor of labor law at the Faculty of Law at USP. Author, among other books, of Moral damage in employment relationships (Studio editors) []

the earth is round there is thanks to our readers and supporters.
Help us keep this idea going.

See this link for all articles