Saga adds source attribution to generative AI videos

Researchers have introduced Saga, a framework that adds source attribution to generative AI videos. Saga can detect which AI system produced a given video and embed provenance metadata. The approach combines visual analysis with watermark detection to trace video origins. Experiments show that Saga reliably identifies the generating system across multiple datasets. The paper, posted on arXiv, discusses potential applications in content verification and copyright enforcement. Developers could integrate Saga into video platforms to combat misinformation. The authors note limitations when faced with heavily edited or compressed footage. Future work will explore broader AI‑generated media and improve robustness against adversarial attacks.