The AI Event: An Ontological Break with Rodowick's Digital Event —A Turn in the Ontology of the Moving Image

The emergence of AI text-to-video generation challenges the foundational premise of film-image theory. Generative models such as Sora and Runway produce footage of photographic precision with no camera and no subject—that is, with no profilmic event. This paper proposes a new concept, the AI Event, to theorize this condition, and argues that AI-generated moving images mark not an extension of D.N. Rodowick's Digital Event but an ontological break from it. The argument proceeds in three stages. First, Rodowick's concept of the Digital Event, however much it stresses discreteness and modularity, implicitly presupposes a moment of capture—that the data to be transformed came from somewhere. Second, by way of Clément Rosset's concept of the double, AI-generated images fall outside the existing taxonomy of images: a double presupposes an original, and an image generated from a statistical distribution can be the double of no single original. Third, the AI Event is defined by four components—Ungraspability of Intermediate Variables, Profilmic Void, Image Without Trace, and Latent Occurrence. The paper rejects, on four grounds, the objection that AI images retain an indirect indexicality (Latent Indexicality) through their training data, and argues that the AI Event inaugurates a post-indexical ontology of the image. The genealogy of image ontology runs: ça a été (Barthes) → ça a été calculé (Rodowick) → ça a été estimé (AI Event).