Imagine asking an AI chatbot which candidate to vote for, only to receive advice that’s as unreliable as a broken compass. This isn’t a hypothetical scenario—it’s the reality uncovered by a recent study on AI’s role in Hungary’s elections. The findings are a wake-up call for anyone who thinks algorithms can be trusted to guide democratic decisions. Personally, I think this issue cuts deeper than most realize. It’s not just about tech failing; it’s about the erosion of trust in systems we’ve started to depend on without scrutiny.
Let’s start with the elephant in the room: AI chatbots like ChatGPT and Gemini are being asked to provide political advice, and they’re doing it poorly. A study by Liberties, a civil liberties group, found that these systems often recommended the wrong parties, missed the correct ones, or even suggested parties that didn’t exist in the election. What makes this particularly fascinating is how confidently the AI delivers its recommendations. They sound authoritative, yet their logic is a black box. In my opinion, this creates a dangerous illusion of reliability. Users are left with the impression that these systems are neutral arbiters of truth, when in reality, they’re built on data that’s inherently biased and outdated.
Take Hungary’s recent election, where the opposition Tisza party won a historic victory. The study revealed a staggering 90% failure rate for ChatGPT when it came to recommending Tisza to users whose profiles aligned with the party. Meanwhile, Fidesz, the ruling party, was correctly identified in about half the cases. This isn’t just a technical glitch—it’s a systemic flaw. What many people don’t realize is that AI models like these are trained on static data. Tisza’s rise to prominence after 2024 meant the models couldn’t ‘see’ it, leaving voters in the dark. This raises a deeper question: Can we trust AI to keep up with the pace of political change, or are we setting ourselves up for repeated failures?
The disclaimers AI systems use—like ‘I cannot give political advice’—are performative. They’re followed by paragraphs of detailed, persuasive analysis that makes users feel they’re getting expert guidance. A detail that I find especially interesting is how these disclaimers coexist with the AI’s willingness to provide recommendations. It’s like a magician who admits they’re not a real wizard but then proceeds to pull a rabbit out of a hat. This contradiction is a ticking time bomb for democracy. If voters start relying on opaque systems that claim neutrality while delivering unverified advice, we’re essentially handing over our political agency to algorithms we can’t hold accountable.
What this really suggests is a regulatory vacuum. The EU’s AI Act and Digital Services Act are supposed to address risks, but they’re not keeping pace with the realities of AI’s political influence. The researchers argue that AI providers should stop offering personalized voting advice unless they can guarantee transparency. I couldn’t agree more. The problem isn’t just the AI itself—it’s the lack of oversight. We’re in a race between technological advancement and the ability to govern it responsibly. If we don’t close this gap, we risk normalizing a system where decisions are made by machines that can’t explain their reasoning, let alone be corrected.
Looking ahead, this isn’t just a Hungarian issue. As AI becomes more integrated into political processes globally, the lessons here are universal. Imagine a future where AI-driven campaigns manipulate voter behavior through micro-targeted advice, or where election outcomes are subtly influenced by algorithmic biases. The implications are staggering. What this study shows is that we’re at a crossroads. We can either treat AI as a tool that needs to be tamed with regulation, or we can let it evolve unchecked, risking the very foundations of democratic participation. The choice is ours—but the clock is ticking.
