Researchers at the University of California, Riverside, in collaboration with YouTube and Google DeepMind, have unveiled SAGA, a forensic framework designed to trace AI-generated videos back to their specific creation systems. As generative artificial intelligence rapidly advances, distinguishing synthetic media from authentic footage grows increasingly challenging. While traditional detection tools establish whether a video is fabricated, SAGA addresses the more complex requirement of source attribution, identifying the exact model and development team responsible for generating the content. Led by doctoral student Rohit Kundu under the supervision of Professor Amit Roy-Chowdhury, the SAGA team developed a methodology that moves beyond single-frame analysis. Video generators leave unintentional visual artifacts that evolve across frames, creating distinctive temporal signatures. By examining both spatial details within individual frames and temporal relationships throughout a sequence, the framework captures how visual elements shift over time. The researchers utilize a technique called Temporal Attention Signatures, or T-Sigs, which averages these patterns across thousands of outputs to generate a unique fingerprint for each AI system. The framework was rigorously tested using public datasets comprising videos produced by nineteen distinct generative models, encompassing both text-to-video and image-to-video architectures. SAGA successfully classified content as real or synthetic, determined the input modality used, differentiated between model versions, and pinpointed the underlying development team. These capabilities represent a significant advancement in digital forensics, where understanding the origin of synthetic media is now considered as critical as verifying its authenticity. The development of SAGA responds directly to escalating concerns regarding the misuse of AI-generated video in misinformation campaigns, corporate fraud, and political manipulation. Entertainment, advertising, education, and social platforms increasingly leverage synthetic media, necessitating robust transparency mechanisms. Regulatory bodies and technology companies require reliable tools to enforce disclosure standards and track the dissemination of synthetic content online. SAGA provides a forensic pathway to meet these demands, enabling investigators to map misinformation networks and hold content generators accountable. This initiative builds upon prior research by the same team, which previously developed AI models for detecting video tampering. As generative AI capabilities continue to outpace detection methods, experts describe the field as a continuous competition. The publication of the SAGA framework underscores a broader industry shift toward proactive accountability. By making the origins of synthetic videos identifiable, the research aims to establish verifiable standards for AI media transparency and mitigate the risks associated with increasingly sophisticated digital deception.