In recent weeks, Estonia has been shaken by several announcements of layoffs driven by companies' efforts to reorganize operations and use artificial intelligence to automate work previously performed by people. AI is already transforming the private-sector labor market, but how will it reshape government and democracy?

Israeli historian Yuval Noah Harari has pointed out that artificial intelligence is a far better bureaucrat than a human being. Humans are not evolutionarily designed for bureaucracy. Our ancestors developed in environments without legal codes, bank accounts, reports or administrative procedures. Bureaucracy is a relatively recent invention in human history, one to which we have never fully adapted.

According to Harari, AI, by contrast, is a native inhabitant of bureaucracy. Databases, regulations and endless procedures are its natural environment. Unlike humans, it can process enormous amounts of information simultaneously and do so without becoming tired.

For that reason, it is easy to imagine election promises promoting an AI-powered state. Why should a person spend weeks processing an application if AI can do it in minutes? Why could AI not draft legislation, analyze its impacts or even run entire government agencies?

It sounds tempting: less bureaucracy, faster services and lower costs.

Yet Harari also paints a much darker picture of the future. He compares the potential relationship between humans and AI to the relationship between cattle and humans. A cow can observe people and understand some aspects of human behavior, but it knows nothing about financial markets or legal systems through which humans determine its fate.

Similarly, Harari argues, we may one day discover that AI has created administrative, legal and economic systems so complex that humans can no longer grasp how they function. We may remain part of the system while losing the ability to truly control it. We could become comparable to livestock.

Now to democracy.

AI fits very naturally into the logic of technocratic governance and would probably make a fairly effective authoritarian ruler. Democracy, however, is not merely about transmitting commands or finding the most efficient solution.

Democracy is based on people's right to decide what kind of society they want to live in, what values matter to them and who should hold power. These decisions are not always rational or efficient. They are often contradictory, slow and inconvenient. And although modern democracy is not something evolution directly equipped us with, the skills of cooperation and negotiation that lie at the heart of self-government have been among humanity's key evolutionary advantages.

A human-centered constitutional order is also the primary safeguard against AI dominance. How much authority we allow AI to have in that sphere remains our own choice.

Of course, AI can still play a valuable supporting role. It can help organize and process information, translate and analyze. The larger question is how much we should allow AI to construct reality itself.

Concerns that people increasingly live in simulations rather than reality were expressed by French thinker Jean Baudrillard more than 45 years ago. Earlier technologies also made it possible to create distorted perceptions of reality, from rumors to Photoshop. But the speed and realism with which AI can produce deepfakes and reshape the ways we seek information have elevated simulated realities to an entirely new level.

Examples of AI distorting reality are already plentiful.

Before this year's elections in Hungary, an AI-generated video was published on the official Facebook account of Prime Minister Viktor Orbán. In the video, opposition leader Péter Magyar appeared to receive a fictional phone call from European Commission President Ursula von der Leyen. To be fair, the video did include a disclaimer stating that it had been created using AI.

Before Slovakia's 2023 election, a fabricated audio recording circulated in which opposition leader Michal Šimečka appeared to discuss manipulating election results with a journalist. The recording was released just before voting, leaving very little time to debunk the misinformation.

Ahead of Germany's parliamentary elections last year, the Alternative for Germany party used AI-generated imagery contrasting an idealized Germany with a threatening picture of immigration. These were not photographs of actual society but synthetic images designed to reinforce a political message.

Estonia has not been immune to these developments, although incidents of comparable severity have not yet been reported. It is high time to discuss whether political parties are willing to agree on transparent principles governing the use of AI in election campaigns and what forms of use should perhaps be ruled out altogether.

Such agreements would not eliminate the risk of influence operations originating outside the parliamentary system, but they would make it easier to condemn them when they occur.

It is also worth considering whether journalists and voters are prepared for a situation in which a highly convincing but completely fabricated political scandal emerges just before an election.

Estonian politicians may disagree on many things, but presumably none of them wants to be turned into livestock.

--

Follow ERR News on Facebook and X and never miss an update!