AI and Democracy: The Risks of Information Manipulation

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 has transformed the way we consume information, but it has also created powerful tools of manipulation that threaten the pillars of modern democracy.

The Current Landscape of AI-Fueled Disinformation

We live in an era in which artificial intelligence algorithms can generate text, images, audio, and video so realistic that it is hard to tell authentic content from fabricated content. When used maliciously, this capability poses a direct threat to the integrity of democratic processes. This article examines how these technologies can be turned into tools for manipulating public opinion, and what legal and regulatory mechanisms are being developed to meet the challenge.

Information manipulation is not new. Political propaganda, fake news, and disinformation campaigns have existed for centuries. Artificial intelligence, however, has given these practices an unprecedented scale and sophistication. Natural language generation tools can produce thousands of persuasive texts in a matter of minutes, adjusting tone, vocabulary, and even ideological slant to fit the target audience. Likewise, deepfake technology makes it possible to create videos in which public figures appear to say things they never said, eroding trust in institutions and in audiovisual records themselves.

The impact of these tools is especially acute during election periods. Automated disinformation campaigns can flood social media with false narratives, creating the impression of consensus where none exists and polarizing public debate. Such content spreads far faster than fact-checking agencies, or voters themselves, can verify it.

This subject is directly related to algorithms in judicial decisions, a topic we cover in a dedicated article.

Deepfakes, Bots, and the Erosion of Public Trust

Deepfakes are perhaps the most visible and disturbing form of AI-driven information manipulation. They are audiovisual content generated or altered by neural networks that can faithfully reproduce anyone’s appearance, voice, and gestures. In politics, this technology can be used to fabricate damaging statements, simulate scandals, or create fictitious situations involving candidates and public officials. The devastating effect is not limited to the immediate harm done by the lie: even after it is debunked, the false content will already have done its work of sowing doubt and distrust.

Beyond deepfakes, networks of bots (automated social media profiles) work in coordination to amplify certain narratives, attack political opponents, and create artificial trends. These profiles, increasingly sophisticated thanks to AI, can mimic human behavior with remarkable precision, making it harder for digital platforms to identify and remove them. The result is a contaminated information environment in which ordinary citizens find it increasingly difficult to tell genuine opinions from orchestrated campaigns.

The greatest threat AI poses to democracy is not the lie it creates, but the doubt it plants about every truth.

This erosion of public trust has consequences that reach beyond the election cycle. When people come to distrust all information as a matter of course (a phenomenon known as the “liar’s dividend”), even true and important news can be dismissed as potentially false. This widespread skepticism weakens democratic debate, makes consensus harder to build, and undermines the legitimacy of institutions.

The Regulatory Framework Under Construction: Brazil and International Experiences

In light of these risks, lawmakers in many countries have sought to establish regulatory frameworks that address AI-driven information manipulation. In Brazil, the legislative debate on regulating artificial intelligence has moved forward through bills that propose, among other measures, mandatory labeling of AI-generated content, platform liability for the spread of deepfakes, and mechanisms for algorithmic transparency.

Brazilian election regulation has already incorporated some relevant provisions. The Superior Electoral Court (Tribunal Superior Eleitoral) has issued resolutions on the use of artificial intelligence in election campaigns, restricting the use of deepfakes and requiring that content produced with AI tools be identified as such. These rules are an important step forward, although their effectiveness depends on enforcement capacity and on the cooperation of digital platforms.

Internationally, the European Union stands out with the AI Act (the European Artificial Intelligence Regulation), which classifies AI systems by risk level and sets specific requirements for applications considered high-risk, including those that may affect democratic processes. The regulation imposes transparency obligations on AI systems that generate or manipulate content, requiring that users be told when they are interacting with artificially produced content.

The United States, for its part, has taken a more fragmented approach, with state laws addressing specific issues (such as criminalizing election deepfakes in some states) and sector-specific regulatory initiatives. The lack of comprehensive federal legislation, however, leaves gaps that bad actors can exploit.

Technical and Legal Challenges in Detection and Accountability

One of the greatest challenges in combating AI-driven information manipulation is the technical difficulty of detecting artificially generated content. Tools exist to detect deepfakes and AI-written text, but they are locked in a constant arms race: every advance in detection methods is met by more sophisticated generation techniques that slip past existing filters. This dynamic makes it unlikely that the solution will be purely technological.

From a legal standpoint, holding anyone accountable for the harm caused by information manipulation faces considerable obstacles. Identifying the authors of disinformation campaigns is often made harder by the anonymity the internet provides and by technological infrastructure spread across multiple jurisdictions. In addition, defining precisely what counts as “information manipulation” (as opposed to legitimate freedom of expression and political humor) calls for careful criteria, so that regulation does not become an instrument of censorship.

The responsibility of digital platforms is central to this debate. The large technology companies that host and distribute content play a crucial role in amplifying (or containing) manipulated material. Deciding how much responsibility these platforms should bear, somewhere between the neutrality of a mere intermediary and shared responsibility for the content they spread, is one of the most contentious points in the regulation now taking shape.

The right to truthful information, although not expressly named as such in the Federal Constitution, can be derived from a systematic reading of fundamental rights, especially freedom of expression, the right to information, and the democratic principle. By distorting the information environment in which citizens form their opinions and make political decisions, AI-driven information manipulation strikes at the core of these fundamental rights.

Paths to Protecting Democracy in the Age of AI

Confronting AI-driven information manipulation requires a multifaceted approach that combines regulatory tools, technological solutions, digital education, and stronger institutions. No single one of these fronts will be enough to contain the problem. Effectiveness depends on coordinated action by government, digital platforms, civil society, and citizens themselves.

On the regulatory front, we consider it essential that Brazilian legislation on artificial intelligence include specific provisions on information manipulation: mandatory labeling of AI-generated content (watermarking), fast-track mechanisms for removing deepfakes during election periods, and clearly defined responsibilities for developers, distributors, and platforms. At the same time, this regulation must protect freedom of expression and avoid creating mechanisms that could be used to silence legitimate voices.

Digital education, often called media literacy, is another indispensable front. Citizens who can spot manipulation techniques, critically evaluate the information they receive, and understand the basics of how the algorithms that mediate their access to information work are the first line of defense against disinformation campaigns. Educational programs of this kind should be built into school curricula and extended to adults through public awareness campaigns.

Finally, strengthening democratic institutions (including the electoral courts, oversight bodies, and the professional press) is essential to provide effective checks on information manipulation. Robust institutions with adequate technical and human resources are better able to identify, investigate, and hold accountable those who spread disinformation, preserving the integrity of the democratic process.

We conclude that the challenge of AI-driven information manipulation is, ultimately, a challenge for civilization. How contemporary democracies respond to this threat will largely define the quality of public debate and the legitimacy of collective choices in the coming decades. Building effective responses demands urgency, but also prudence, so that the cure does not turn out to be worse than the disease.

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