In response to epistemic challenges posted to the public by hyper-realistic characters in AI-generated synthetic media, this chapter aims to answer the question: to what extent is generative AI responsible for its generated texts, images, audio, or videos, especially in situations involving deepfakes? I argue that generative AI models such as an artificially intelligent speech generator or a Generative Adversarial Network (GAN) should be accountable for the transparency and truthfulness of its generated output due to its practical impacts on the public, particularly considering the injustice and potential epistemic threats posed by political deepfakes. The latest trend in AI regulations, as presented by the 2024-implemented EU Artificial Intelligence Act (AIA) and the 2024-signed California AI Transparency Act, shows the move toward a more comprehensive regulatory framework for AI, which aligns with my argument on AI accountability.