Researchers have built a tool that can identify the AI used to make a fake video

UC Riverside researchers, with YouTube and Google DeepMind, developed SAGA to detect AI-generated videos and trace them back to the specific model or development team.

The system looks for subtle spatial and temporal patterns, including Temporal Attention Signatures, that reveal a video’s synthetic origin.

It was tested across 19 video generators, covering both text-to-video and image-to-video systems.

The work could improve misinformation tracking, regulatory oversight, and accountability for AI video content.