Artificial intelligence may not only produce deepfakes and political disinformation but also influence voters’ decisions at the ballot box. – It may turn out that it is not political parties but algorithms that will decide who wins the next elections – warns Jakub Szymik, founder and president of the Digital Democracy Observatory Foundation.
During the UN Global Dialogue on AI governance, a strong warning was issued: AI may influence elections and disrupt democratic processes. According to Szymik, AI models can adapt to specific voters, conversations and user needs, potentially tipping the scales in one direction or another.
The foundation’s report „Droga do karty wyborczej” (The Road to the Ballot) highlights that voluntary commitments from tech companies – such as the 2024 Munich Declaration – are not enough. While major players like OpenAI, Anthropic and Google have begun publishing policies on transparency of election-related content in the context of US midterms, the problem of AI-generated misinformation is global.
In the UK, for example, AI models were found to hallucinate election dates, party programs and even the existence of candidates. During the European Parliament elections, thousands of deepfake posts appeared showing Ursula von der Leyen or Stalin dancing on the grave of the EU. Szymik expects a similar wave of AI-generated content during upcoming Polish elections.
Political chatbots and virtual influencers
Election staffs will likely use chatbots to encourage people to vote. Entirely AI-generated influencers may appear, sometimes without proper labeling. Even if labeled, voters might ignore the tags. Szymik notes that Grok, the chatbot on X, openly declared it would vote for the Confederation party – a curious stance given that Meta removed the party’s Facebook page in 2022. Today, Confederation leads in social media reach thanks to polarizing language and viral posts by figures like Ewa Zajączkowska-Hernik.
– It can happen that algorithms, not political parties, decide the outcome of the next election – warned Szymik.
Polish legislative response
The foundation proposes concrete solutions: mandatory labeling of AI-generated content, auditing of recommendation algorithms, and a ban on deepfakes impersonating candidates. The EU’s AI Act already requires labeling AI-produced material. Brazil has issued clear guidelines for AI in elections since 2024. Poland’s Senate has prepared a bill that would allow imposing fines of up to 6 percent of global annual revenue on major platforms like Meta, Google, TikTok and X for illegal, unlabeled political advertising.
The bill would empower multiple institutions – the National Electoral Commission, the Personal Data Protection Office, the National Broadcasting Council, the Office of Electronic Communications and the Office of Competition and Consumer Protection – to enforce the law. However, Szymik fears these bodies may clash over jurisdiction. He also warns that the definition of political advertising could inadvertently cover NGOs, interest groups and climate organizations.
– We encourage the parliament and the presidential palace to push these regulations through together, understanding the gravity of the situation – added Szymik.
Źródło: WNP.PL, Fot. Shutterstock (dee karen), mat. Fundacji Obserwatorium Demokracji Cyfrowej
