YouTube to begin automatically labeling AI videos (arstechnica.com)

YouTube is set to enhance its approach to identifying AI-generated videos by implementing more prominent automatic labeling. Following a rather ineffective initial attempt in 2024, where AI content often revealed itself through its odd characteristics, the platform acknowledges the growing realism achieved by advanced AI models like Seedance, Runway, and Google’s Veo. As a response, YouTube will now employ “new internal signals” to flag videos that exhibit significant photorealistic AI use, moving away from its reliance on creators voluntarily disclosing their use of AI tools. This development is significant for the AI/ML community as it addresses the increasing challenge of discerning AI-generated content from genuine human creation—an issue that impacts everything from media integrity to content regulation. While the specifics of the detection signals remain unclear, notable triggers include C2PA metadata indicating an AI source and watermarking from Google's tools. Creators will still need to declare their use of AI when uploading, but the new automatic flagging introduces a layer of accountability aimed at curbing potential misrepresentation in content creation.