This conceptual legal-policy position paper develops the democratic chain-of-harm framing for electoral deepfakes and platform responsibility in EU law. The paper argues that electoral deepfakes should not be understood merely as isolated instances of false content requiring rapid removal. Instead, they are conceptualised as terminal manifestations of a broader democratic chain of harm involving generative AI systems, campaign incentives, platform dependency, recommender systems, microtargeting, algorithmic amplification and delayed legal response. Building on the chain-of-harm framework previously developed in relation to crypto-infrastructures, the paper extends the concept of systemic responsibility to democratic information infrastructures. It situates the argument within selected EU legal instruments, including Article 114 TFEU, the AI Act, the Digital Services Act and Regulation (EU) 2024/900 on the transparency and targeting of political advertising. The paper emphasises that rapid takedown mechanisms are necessary but insufficient where democratic harm is produced through speed, virality and timing. It proposes a rights-sensitive and procedurally accountable approach to anticipatory governance, including safeguards for satire, parody, journalism, artistic expression and lawful political speech. This version is a conceptual legal-policy position paper. It does not provide an empirical study of electoral deepfake prevalence, test detection tools, or offer a complete doctrinal analysis of all applicable EU, national electoral, media or constitutional-law provisions.
Ian Ardaseer (Sun,) studied this question.