Protecting democracy: comparing codes of conduct for generative AI use during elections

2024 has been called the ‘super election year’ with more than four billion citizens participating in national elections in 74 countries. The introduction of Generative Artificial Intelligence was deemed to ‘upend global elections’ by experts and commentators, particularly through the spread of manipulated audio, images, text or videos. These anxieties were prominently voiced at international fora, including the World Economic Forum and the Munich Security Conference in 2024. To address these risks and protect the integrity of elections, some countries introduced guidelines or codes of conduct on AI use during elections, often swiftly. These codes were introduced by political actors or national election commissions, and they specify either voluntary guidelines or mandatory behavior. Through a comparative analysis of national and international codes of conduct, guidelines, and legal text introduced in 11 countries in 2024, this paper argues that emerging codes of conduct for the use of generative AI in elections reflect competing interpretations of what constitutes a threat to democratic integrity, with particular emphasis on safeguarding public trust rather than solely preventing direct manipulation of outcomes. Through a comparative analysis of regulatory approaches and codes of conduct across political systems, we explore how these instruments define and delimit harmful uses of generative AI and how they institutionalize responsibility through varied models. Our overall assessment is that while these codes represent important early attempts at governance, they often lack enforceability, consistency, and normative clarity, leaving significant gaps in the regulation of generative AI use during elections.