Beyond the Liar's Dividend: AI Political Deepfakes, Cognitive Sovereignty, and the Imperative for Infrastructure Accountability

Generative artificial intelligence is now being used to spread political disinformation, and at its base, this technology is also undermining cognitive sovereignty—the right of the people to form opinions based on facts. With the spread of hyper-realistic deepfakes on social media, we have reached the "liar's dividend" phenomenon; genuine evidence of political crimes can be dismissed as "AI-generated fake news". Therefore, the current intervention measures by platforms, such as warning labels, are insufficient and have shifted the responsibility for verification to ordinary citizens. Using the dual-theoretical system of deontology and utilitarianism, this paper argues that AI political deception is inherently immoral because it violates procedural justice and is thus harmful in practice due to the damage to democratic trust. In light of the problem of regulatory abuse and censorship, this paper proposes a new model of infrastructure accountability for content moderation. In line with the idea of information fiduciary duty, platforms will be required to increase algorithmic transparency and introduce a tiered regulatory penalty system to protect the rule of law and freedom of expression.