Netherlands: Data Protection Authority issued a report examining risks of using AI chatbots as voting aids for elections

On 21 October 2025, the Netherlands Data Protection Authority (DPA) issued a report examining the risks of using AI chatbots as voting aids for elections. The DPA found that chatbots provide strongly biased and polarised political advice, misrepresenting the Dutch fifteen-party system. In an experiment testing four chatbots with fictional voter profiles based on Kieskompas and StemWijzer data, two parties, GroenLinks-PvdA and PVV, dominated first-place recommendations, while parties such as CDA, ChristenUnie, DENK, Forum voor Democratie, and SGP were rarely suggested. The study identified that they are funneling left-leaning voters toward GroenLinks-PvdA and right-leaning voters toward PVV, oversimplifying and polarising the political landscape. The DPA noted that, unlike traditional voting aids, chatbots lack transparency, neutrality, and verifiability, with underlying models containing opaque biases. Despite safeguards claimed by providers, the DPA found chatbots offered voting advice in over 99.9% of 21,000 test queries. The DPA also warns citizens against using chatbots for voting advice and urges developers to implement effective safeguards, noting that AI systems influencing elections are classified as high-risk under the EU AI Regulation.