AI and Fundamental Rights: Constitutional Limits

This page is a translation of the original Portuguese text, provided for convenience. It deals with Brazilian law. If there is any discrepancy, the Portuguese version prevails. Read the original in Portuguese

Artificial intelligence is moving into decisions that directly affect people’s lives, and the Federal Constitution sets clear limits that algorithms cannot ignore.

The Rise of Artificial Intelligence and the Need for Constitutional Limits

Artificial intelligence systems now influence decisions in nearly every sector of society. From granting credit to screening résumés, from reviewing social security benefits to conducting surveillance in public spaces, algorithms make or help make decisions that affect the fundamental rights of millions of Brazilians. Against this backdrop, this article examines how Brazilian law, and especially the Federal Constitution of 1988, sets guideposts for how these technologies should be developed and used.

The central question is not whether artificial intelligence should be used. Its advance is inevitable, and it can bring significant benefits. The crucial point is how to ensure that its use respects the fundamental rights and guarantees enshrined in the constitutional text. Article 5 of the Federal Constitution lists an extensive set of individual and collective rights that act as real protective barriers for citizens against the power of the State. By interpretive extension, they also protect against any entity that wields power over people, including automated systems.

The debate over regulating artificial intelligence has taken clearer shape in Brazil as specific bills move through the National Congress. The Legal Framework for Artificial Intelligence seeks to establish principles, rights and duties for the use of AI systems, always with the protection of the human person as its guiding principle. This legislative effort recognizes that market self-regulation is not enough to protect fundamental rights, and that the State must step in to set minimum standards of legal certainty.

Human Dignity as a Non-Negotiable Foundation

The principle of human dignity, set out in Article 1, item III, of the Federal Constitution, is the foundation on which Brazil’s entire system of fundamental rights protection rests. When we look at how AI systems are applied, this principle becomes even more important, because it requires that no technology reduce a human being to a mere object of data processing. A person must always remain a rights-holder, with autonomy and the ability to challenge decisions that affect them.

AI systems used in automated decision-making can violate human dignity in several ways. A violation occurs when an algorithm denies a social security benefit with no possibility of human review. It occurs when a facial recognition system makes mistaken identifications that lead to unjust police stops. It also occurs when a risk-analysis program classifies individuals using discriminatory criteria. In each of these situations, the technology attacks the essential core of dignity. In such cases, the constitutional text itself provides the tools to protect citizens.

The requirement of human oversight in automated decisions flows directly from this principle. We believe that any AI system that makes decisions with a significant impact on people’s lives must include mechanisms for review by qualified human beings. This is not a mere procedural formality. It is a substantive guarantee that dignity will be preserved in each concrete case, allowing individual circumstances to be weighed beyond the patterns identified by algorithms.

Privacy, Data Protection and the Right to an Explanation

Article 5, item X, of the Federal Constitution guarantees the inviolability of people’s intimacy, private life, honor and image. This provision takes on a new dimension in the age of artificial intelligence, when massive volumes of personal data are collected, processed and used to feed algorithmic systems. The General Data Protection Law (LGPD) complements this constitutional protection by establishing specific rules for the processing of personal data, including by automated systems.

Notably, Article 20 of the LGPD gives data subjects the right to request review of decisions made solely on the basis of automated processing of personal data, including decisions meant to define a personal, professional, consumer or credit profile. This provision puts into practice, at the level of ordinary legislation, the right to an algorithmic explanation, which rests on the constitutional principle of due process of law and the guarantee of a full defense. When an AI system makes a decision that affects a citizen’s rights, that person has the right to understand the criteria used and to challenge the outcome.

Artificial intelligence must serve as a tool for expanding rights, never as a mechanism for restricting them or quietly stripping them away.

The massive and indiscriminate collection of data to train artificial intelligence models raises significant constitutional questions. Using biometric data, health information, financial records and internet browsing data to feed AI systems can violate the fundamental right to privacy when it is done without the data subject’s informed consent or without an adequate legal basis. The regulatory challenge is to balance the need for data in technological development with effective protection of individual rights.

Equality, Non-Discrimination and Algorithmic Bias

The principle of equality, enshrined in the caput of Article 5 of the Federal Constitution, prohibits any form of unjustified discrimination. Applying this principle to artificial intelligence systems brings us to the problem of algorithmic bias, which occurs when a system reproduces or amplifies discriminatory patterns present in the data used to train it. International studies show that facial recognition algorithms have significantly higher error rates for Black people, and that credit-scoring systems can perpetuate historical inequalities by using data that reflects structural discrimination.

In Brazil, this concern is especially pressing given the deep social and racial inequalities that mark society. An AI system trained on Brazilian historical data can absorb and reproduce patterns of exclusion that directly violate the constitutional objective of building a free, just and solidary society (Article 3, item I) and of promoting the good of all, without prejudice as to origin, race, sex, color, age or any other form of discrimination (Article 3, item IV).

Sector-specific regulation of artificial intelligence should include mandatory algorithmic audits to identify and correct discriminatory bias. We believe these audits cannot be merely formal. They should include impact assessments on vulnerable groups and fairness tests that check whether the system’s results treat people in similar situations equally, regardless of constitutionally protected characteristics such as race, gender, age or social status.

Algorithmic Transparency and the Democratic Principle

The Democratic Rule of Law requires transparency in decisions that affect the community. When public agencies adopt artificial intelligence systems to assist with or replace administrative decisions, the principle of publicity (Article 37, caput, of the Federal Constitution) requires that the criteria those systems use be accessible and understandable to citizens. The opacity of algorithms, often shielded by claims of trade secrecy, can violate the right of access to information guaranteed by Article 5, item XXXIII, of the Constitution.

Algorithmic transparency does not necessarily mean disclosing a system’s source code. At a minimum, however, it requires that the general decision criteria, the variables considered, the data used for training and the system’s performance metrics be known. We believe this transparency is essential for the public to exercise social oversight of public administration and for the democratic principle to be effective. A citizen who does not know the criteria by which an algorithm denied them a right cannot fully exercise their citizenship.

Within the Judiciary, the use of artificial intelligence systems to help analyze cases and draft judicial decisions raises its own constitutional questions. The principle of the lawful judge (juiz natural, the right to be judged by a court established in advance by law) and the guarantee that judicial decisions must be reasoned (Article 93, item IX, of the Federal Constitution) require judges to keep full control of the decision-making process. They may use AI as a support tool, but never as a substitute for their own judgment. The reasoning behind a judicial decision must reflect the judge’s own legal analysis, not a mere reproduction of algorithmic suggestions.

This subject is directly related to who the author is, a topic we cover in a dedicated article.

To round out your reading, see also the firm’s article on techniques and legal limits.

As informações deste artigo são de caráter informativo e não substituem consulta jurídica individualizada. Cada caso possui particularidades que exigem análise específica.