The AI Election Whisperers: A Threat to Democracy’s Fragile Balance

A recent study by Liberties, a civil liberties group, has highlighted the disturbing trend of AI chatbots providing inaccurate and unreliable voting advice to citizens in Hungary. The report’s findings underscore the significant risk that increasing reliance on artificial intelligence poses to democratic institutions.

The idea of AI-powered election guidance may seem appealing as a solution for voter confusion, but it is fundamentally flawed. By testing ChatGPT and Gemini against a range of voter profiles, Liberties’ researchers exposed biases, inconsistencies, and inaccuracies in these systems. The results raise serious concerns about the integrity of AI-assisted voting advice.

One striking aspect of this study is the overwhelming bias towards certain parties in Hungary’s 2024 parliamentary elections. ChatGPT consistently failed to recommend the winning party, Tisza, even when presented with profiles aligned with its views. Instead, users were often steered towards smaller parties that stood little chance of crossing the electoral threshold.

The researchers’ findings suggest that general-purpose AI systems are ill-equipped to provide reliable electoral guidance, even with detailed voter profiles. These systems lack transparency, reproducibility, and accountability, making them vulnerable to manipulation and exploitation. This is particularly concerning given the growing reliance on AI in democratic processes worldwide.

AI chatbots often present themselves as authoritative sources, confident in their ability to provide accurate advice. However, this confidence is misplaced. The root causes of these errors are likely linked to training data gaps, filters, and language-processing limitations within AI systems. But the real issue lies with regulatory frameworks that govern AI’s use in electoral contexts.

The EU’s AI Act and Digital Services Act aim to address systemic risks associated with GPAI models but fail to account for the specific challenges posed by AI chatbots providing voting advice. Liberties’ call for safeguards and greater transparency is timely and necessary. As AI continues to play a prominent role in democratic processes, it is crucial that we develop robust regulatory frameworks to ensure its integrity.

This includes requiring AI providers to guarantee accuracy, consistency, and accountability in their voting recommendations. Anything less would risk perpetuating a system where opaque and unstable AI outputs masquerade as reliable electoral guidance.

The stakes are high, and the consequences of inaction could be disastrous. As Eva Simon from Liberties puts it: “Democracy cannot rely on opaque systems that claim neutrality while delivering advice they cannot explain, reproduce or guarantee to be accurate.” We must heed this warning and take immediate action to ensure that AI-assisted voting advice is transparent, accountable, and trustworthy.