Generative AI is transforming the media ecosystem and, with it, the challenge of verifying authenticity. This seminar examines the shift from standalone deepfake detectors toward AI forensic agents: systems that can orchestrate multiple detectors, reason over provenance and context, express calibrated uncertainty, and support human analysts in high-stakes decisions. Drawing on Truebees' experience in deepfake detection and recent research on explainable, uncertainty-aware AI forensics, the talk will discuss why multimedia authenticity should be treated as a systems problem rather than a single-model classification task. It will highlight practical lessons from deployment, current technical limitations, and emerging agentic approaches that may define the next generation of trustworthy multimedia forensics. The seminar is aimed at researchers, students, and practitioners interested in computer vision, trustworthy AI, and the future of media verification.

The event is organized in the framework of the Euregio Visper Project.