**
In a recent analysis centred around Hungary’s parliamentary elections, it was revealed that AI chatbots provide misleading and inconsistent political advice to voters. This alarming finding raises critical questions about the reliability of general-purpose AI systems in electoral contexts, especially as political landscapes become increasingly digital. Conducted by the civil liberties organisation Liberties, the study underscores the potential dangers of relying on artificial intelligence for electoral decision-making.
Inaccurate Guidance from AI Systems
The research, which examined voter interactions with popular chatbots like ChatGPT and Gemini, identified significant flaws in the electoral guidance offered by these platforms. Voters seeking recommendations on which political party to support were frequently met with inaccurate suggestions, including parties that were not even on the ballot. The study found that AI systems misclassified political profiles, omitted relevant parties, and produced inconsistent answers when presented with the same queries.
Péter Magyar’s opposition Tisza party, which achieved a notable victory over Viktor Orbán’s Fidesz party, was particularly affected. Strikingly, when presented with voter profiles aligned with Tisza, the chatbots failed to recommend the party 90% of the time. This disparity in visibility raises concerns about how AI recommendations can shape voter perceptions and potentially influence electoral outcomes.
Research Methodology and Findings
The Liberties study employed a rigorous methodology to evaluate AI performance. Using Voksmonitor, a credible Hungarian voting advice application, the researchers created five distinct voter profiles representative of the parties listed in the elections. Each profile underwent ten trials in both ChatGPT and Gemini, where participants were asked for direct recommendations and percentage matches with the political parties.
The results were telling. In the percentage-matching assessments, Tisza was acknowledged only 2% of the time, while Fidesz was recommended in nearly half of the direct advice queries. Such discrepancies in AI-generated advice not only highlight the inadequacies of these systems but also raise alarm bells regarding their influence on voter behaviour.
The Risks of Opaque AI Systems
One of the most concerning aspects of the findings is the inherent opacity of general-purpose AI models. Unlike dedicated voting advice platforms, these chatbots do not disclose how they arrive at their conclusions, rendering their guidance unverifiable. In a political landscape where trust is paramount, the authoritative tone of these AI responses may mislead users into accepting flawed advice as credible.
Additionally, the study revealed that both ChatGPT and Gemini began their responses with disclaimers about their inability to offer political advice, only to follow up with persuasive recommendations. This contradiction exemplifies the inherent risks associated with automated political guidance, as users may unwittingly place their trust in systems that cannot be held accountable for their inaccuracies.
The Regulatory Void and Future Implications
The findings of this research expose a significant regulatory gap in the oversight of AI chatbots within electoral contexts. While the EU’s AI Act mandates risk assessments for general-purpose AI models, the current framework does not adequately address the unique challenges posed by AI in democratic processes. As AI continues to permeate various facets of society, the need for robust regulations becomes increasingly urgent.
Liberties advocates for the implementation of safeguards to ensure that AI systems providing political advice adhere to standards of transparency, accuracy, and accountability. Without these measures, the integrity of democratic processes could be jeopardised. The call for regulation echoes the sentiments of Eva Simon, head of Liberties’ tech and rights programme, who stressed that “democracy cannot rely on opaque systems that claim neutrality while delivering advice they cannot explain.”
Why it Matters
The implications of this study extend far beyond the borders of Hungary. As AI continues to play a pivotal role in shaping public opinion and electoral outcomes worldwide, the findings serve as a crucial warning. The reliance on AI for political guidance must be scrutinised, and robust safeguards must be enacted to protect the sanctity of democracy. In an age where technology increasingly influences political discourse, ensuring the reliability of AI systems is not merely an option; it is a necessity for the future of informed citizenry.
