Artificial intelligence and elections: Europe needs more than recommendations

The announcement from the European Commission regarding the development of recommendations for the use of artificial intelligence in electoral processes represents a necessary and welcome step. It is a sign that European institutions are beginning to understand the extent of the transformations brought about by artificial intelligence on the public space, the circulation of information, and the mechanisms through which citizens form their political opinions.

However, there is a risk that this debate may remain confined to a technological logic, while the issue is, in essence, a democratic and societal one.

In recent years, public attention has focused almost exclusively on the most visible phenomena: deepfakes, synthetic images, manipulated videos, automated armies of bots, or coordinated disinformation campaigns. All of these are real threats and must be taken seriously. But they represent only the visible part of a much deeper phenomenon. The real challenge for European democracies is not the existence of false content. The real challenge is the emergence of technological systems capable of organizing, distributing, and amplifying collective attention on an unprecedented scale in human history.

In my opinion, the future of European elections will not be decided primarily by the falsification of reality, but by the invisible management of visibility. It will not matter who produces a message the most, but who controls the mechanisms that give that message access to millions of people. This is why the European debate must shift from the simple issue of content to the problem of informational infrastructures. A fake video can be identified, a manipulated photo can be verified, or a fabricated message can be dismantled.

It is much more difficult to identify the moment when an algorithm decides that a certain emotion, a certain theme, or a certain political actor should receive exceptional, repeated, and disproportionate exposure. Here lies the fundamental vulnerability of contemporary societies, and we must understand that we are not just living in an age of disinformation; we are living in an age of amplification.

Traditionally, the public space was organized through relatively visible institutions: the press, political parties, universities, civic organizations, opinion leaders. Today, a significant part of the public opinion formation process is mediated by automated recommendation systems, which operate according to opaque criteria and are primarily optimized for capturing attention. This transformation fundamentally changes the nature of democratic competition, as we are no longer witnessing just a confrontation between ideas; we are witnessing a competition for algorithmic visibility. Parties, governments, interest groups, external actors, and radical organizations are fighting for access to the infrastructures that distribute public attention.

Artificial intelligence is not just a tool for content production; it becomes a structural actor in the public space and an actor that influences the public agenda, collective priorities, risk perception, and even the definition of social reality. This is why I believe that the Commission's recommendations should be just the first step in a much more ambitious process.

Europe needs a paradigm shift.

First of all, we need to develop independent auditing mechanisms for recommendation and algorithmic amplification systems during electoral periods. Democracy cannot function effectively when the mechanisms that organize the circulation of information are practically impossible to evaluate from the outside. It is not enough to know who pays for a political advertisement; we must also know how it is amplified. Furthermore, it is not enough to identify manipulative content; we must understand why it reached millions of people. We are currently observing the effects, but we need to understand the mechanisms.

Secondly, Europe needs systematic tools for measuring informational vulnerability. Currently, we measure inflation, unemployment, poverty, economic competitiveness, or the level of innovation, but we do not even think about measuring sufficiently the capacity of a society to resist informational manipulation. We should measure the level of informational anomie, the degree of epistemic fragmentation, and trust in legitimate sources of knowledge. Unfortunately, we are not aware that we should measure the cognitive autonomy of citizens. All of these have become strategic variables for the functioning of democracy, and in the absence of such tools, Europe will continue to react to crises after their effects have already become visible.

Thirdly, it is necessary to build a permanent infrastructure for democratic resilience, not just temporary working groups or periodic reports full of recommendations. We need permanent institutional capacities that connect academic research, electoral authorities, regulatory authorities, civil society, and digital platforms.

The European Democracy Shield represents a promising start, but its success will depend on the ability to shift from a defensive approach to a proactive one. Instead of responding exclusively to informational attacks, we need to develop the capacity to identify vulnerabilities before they are exploited.

At the same time, Europe must avoid an important trap. The response to new threats must not lead to excessive limitations on freedom of expression or the transformation of the digital space into a system of generalized surveillance. Democracy cannot be defended by methods that undermine its own values. Therefore, it is essential that transparency, auditability, and accountability of platforms become central principles of European digital governance.

Let it be clear, the fundamental problem is not the existence of artificial intelligence; the problem is the distribution of power in a society where artificial intelligence manages huge volumes of information, shapes attention, and influences collective behaviors. This is why the debate about AI and elections is, in fact, a debate about the future of democracy, not one about technology. We are discussing the capacity of citizens to make autonomous decisions and, above all, about the integrity of public deliberation. The discussion is about trust, legitimacy, and social cohesion.

Europe needs recommendations regarding the use of artificial intelligence in elections, but it also needs something more important: it needs a theory and a policy of democratic resilience adapted to the age of algorithms. Otherwise, we risk regulating the tools and ignoring the profound transformation of the social and cognitive environment in which contemporary democracy operates.

The Romania 2024 scenario, a warning

Romania probably offers the most relevant European case study for this new era of democratic vulnerability. The electoral crisis of 2024 was not just an internal political controversy; it was the first moment in which the European Union faced directly the possibility that an electoral process could be significantly influenced by the combination of digital platforms, recommendation algorithms, coordinated behaviors, and opaque informational operations.

The annulment of the first round of the presidential elections by the Constitutional Court, the subsequent investigations by the European Commission into TikTok, and the debate regarding external interference have turned Romania into a European laboratory for new democratic risks. Beyond political and legal disputes, the Romanian case raises a fundamental question: what exactly needs to be protected in a digital democracy? The outcome of the vote? The fairness of the campaign? Or the informational ecosystem in which citizens form their electoral opinions?

In my opinion, the main lesson from the Romanian case is that democratic institutions are prepared to check the ballots, campaign financing, and electoral procedures, but they are much less prepared to analyze the digital infrastructures that shape collective attention. In Romania, the problem was not exclusively the existence of false information or Russia's hybrid war; the problem was the suspicion that certain distribution and amplification mechanisms produced disproportionate visibility for certain messages and certain candidacies. This is also why European investigations focused not only on the content but also on how recommendation systems work and the possibility of manipulating them.

The Romanian case shows that electoral vulnerability in the 21st century no longer appears only on election day; it begins months in advance, in personalized content flows, in automated recommendation systems, in influencer networks, in digital communities, and in ecosystems that transform attention into a political resource. For the first time in the recent history of the European Union, an electoral crisis has prompted investigations into the role of a digital platform in influencing the democratic process. This should serve as a wake-up call for all of Europe. Therefore, the lesson from Romania is not just about electoral security; it is a lesson about the necessity of building a new generation of democratic institutions capable of understanding and monitoring algorithmic power.

If until now democracies had to defend freedom of expression against censorship, the challenge of the next decade will be the defense of citizens' autonomy in an informational environment shaped by automated influence systems. In this sense, Romania does not represent a European exception; Romania represents a European warning.