Cisco releases open-source toolkit for verifying AI model lineage

Cisco's Model Provenance Kit is an open-source Python toolkit and command-line tool designed to detect whether two transformer models share a common origin by first comparing architecture metadata and then, if needed, extracting five complementary weight-based signals (Embedding Anchor Similarity, Embedding Norm Distribution, Norm Layer Fingerprint, Layer Energy Profile, and Weight-Value Cosine). The report details how the kit operates (compare and scan modes), its benchmark performance on challenging model pairs, deployment characteristics, and its role in addressing AI supply-chain provenance and regulatory/compliance concerns.