AI and Access to Public Information

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 changing how citizens and governments handle access to public information, creating unprecedented opportunities and urgent regulatory challenges in Brazil.

The new landscape of access to public information in the age of artificial intelligence

We are living through a profound shift in the relationship between technology and government transparency. In Brazil, the Access to Information Law (Law 12.527/2011) laid the foundation for public transparency, but it could not have foreseen the impact that artificial intelligence systems would have on how government data is produced, stored, and made available. Today, AI-based tools can process massive volumes of public information, automatically categorize documents, respond to access requests quickly, and even identify patterns in databases that once went unexplored.

Public agencies at many levels of government already use chatbots and virtual assistants to handle citizen requests about government services and information. The Judiciary, for example, has invested in intelligent systems to screen cases, classify petitions, and even suggest draft decisions. These initiatives broaden access to information, but they also raise fundamental questions about the quality, accuracy, and fairness of automated responses.

We must keep in mind that access to public information is a fundamental right under Article 5, item XXXIII, of the Federal Constitution. Any technology that affects this right must be evaluated from a constitutional perspective, so that automation does not become an obstacle to the full exercise of citizenship. When the government adopts AI, it cannot give generic answers that hollow out the right to information, nor can it rely on opaque systems that make it hard to understand the criteria used to provide or deny data.

Regulating AI in Brazil’s public sector: the legal framework and its developments

Brazil is moving toward a regulatory framework for artificial intelligence, with several bills pending in the National Congress. The legislative debate has centered on principles such as algorithmic transparency, accountability, non-discrimination, and human oversight, all of which apply directly to the use of AI in managing public information. Sector-specific regulation is especially important because the government handles sensitive data on millions of Brazilians and makes decisions that affect fundamental rights.

The General Data Protection Law (Law 13.709/2018, known as the LGPD) already provides important benchmarks for the use of AI in the public sector. Article 20 of the LGPD gives data subjects the right to request review of decisions made solely through automated processing of personal data. Applied to access to public information, this means that you have the right to challenge automated responses and to demand that a public employee reevaluate an answer provided by an AI system.

The Brazilian Artificial Intelligence Strategy (EBIA), established by MCTI Ordinance 4.979/2021, also offers relevant guidance. Among its strategic pillars, it stresses the need for responsible AI governance, with special attention to the transparency and explainability of the systems the government uses. These principles are essential to ensure that access to public information is not undermined by the technical complexity of the algorithms in use.

Algorithmic transparency in the public sector is not merely a technical matter. It is a democratic requirement that guarantees citizens the right to understand how automated decisions affect their access to information.

One point deserves special attention: the intersection between the Access to Information Law and AI systems. When a public agency uses an algorithm to decide which information is public and which must be classified as confidential, clear auditing and oversight mechanisms are necessary. Algorithmic opacity cannot serve as a justification for denying access to data that, under current law, should be available to citizens.

Concrete opportunities: how AI can expand public transparency

Artificial intelligence offers powerful tools for expanding access to public information in ways that would be impossible with traditional methods. Natural language processing systems can make government databases more accessible, allowing citizens to run queries in plain language and receive understandable answers, without having to master bureaucratic terminology or know how the agency is organized.

Several promising applications stand out. Automatic classification of public documents can dramatically speed up responses to access-to-information requests, shortening response times and reducing the workload of the employees in charge. OCR (optical character recognition) systems combined with AI can digitize and index historical archives of public documents, making available information that was once confined to physical files that were hard to search.

Predictive analysis also opens interesting possibilities for proactive transparency. Public agencies can use AI to identify which types of information the public requests most often and post them proactively on transparency portals, before citizens ever need to file a request. This forward-looking approach fits squarely with the spirit of the Access to Information Law, which treats proactive transparency as an instrument of democracy.

AI tools can also help detect inconsistencies in public data, supporting public oversight of the government. Algorithms that cross-reference information from different government databases can identify irregularities in bidding processes, contracts, and public spending, strengthening civil society’s role in monitoring the use of public funds.

Risks and challenges: when AI can restrict access to information

Not all of AI’s effects on access to public information are positive. There are concrete risks that regulation and society must address. The first is so-called “algorithmic opacity,” a situation in which the criteria an AI system uses to classify, filter, or prioritize information are not understandable to citizens and, in many cases, not even to the public officials who operate the systems.

When poorly implemented, automated citizen service can create additional barriers to information. Chatbots that fail to understand users’ requests, systems that steer citizens toward standardized answers that do not match their real needs, and digital interfaces that exclude people less familiar with technology are examples of how AI can, paradoxically, make it harder to exercise a fundamental right.

Algorithmic bias is another significant risk. AI systems trained on historical data can reproduce and amplify existing inequalities in access to information. If certain social groups have historically had less access to digital public services, algorithms may be calibrated in ways that perpetuate that exclusion, prioritizing requests from profiles that already enjoy privileged access to government information.

Data security is also a concern. AI systems that process large volumes of public information can become targets of cyberattacks, with serious consequences for citizens’ privacy and the integrity of government data. Concentrating information in automated systems enlarges the attack surface and demands robust investment in cybersecurity, something not every Brazilian public agency is prepared to provide.

Paths toward balanced regulation of AI in access to public information

We believe that regulating AI in the public sector must balance the promotion of innovation with the protection of fundamental rights. The first step is to ensure that any AI system used to manage public information is subject to algorithmic impact assessments: technical documents that analyze the technology’s risks and benefits before it is deployed. These assessments should be public and accessible, allowing scrutiny by civil society and oversight bodies.

We consider it essential to create independent audit mechanisms for AI systems used by the government. Just as government accounts are audited by courts of accounts, the algorithms that affect access to public information should undergo periodic review by independent entities able to assess whether the systems operate in accordance with legal and constitutional principles.

Training public employees is another indispensable element. Many agencies adopt AI technologies without giving their staff the training needed to understand how these systems work, oversee their decisions, and step in when necessary. Digital literacy and AI literacy programs for public servants are strategic investments to ensure that the technology serves the public interest.

Public participation in the regulatory process is equally important. Public consultations, hearings with experts, and citizen feedback mechanisms should inform both the drafting of rules and the deployment of AI systems in public agencies. Regulation cannot be built by technicians and legislators alone, because it must reflect the real needs and concerns of the people who will be affected by these technologies.

Finally, Brazil should look to relevant international experience. The European Union, through its AI Act (Artificial Intelligence Regulation), classified AI systems used by the public sector as “high risk,” subjecting them to strict requirements for transparency, human oversight, and conformity assessment. This model offers valuable lessons for building Brazil’s regulatory framework, provided it is adapted to the particularities of Brazil’s legal system and administrative reality.

This topic is directly related to digital identity and access to, which we cover in a dedicated article.

This topic is directly related to AI and public surveillance, which we cover in a dedicated article.

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