Vanta#199Screen an AI vendor's artifacts against how they say the model was built.
Catches "we trained our own model" when the evidence shows a LoRA adapter or a wrapper. Runs entirely on your machine; a consistency screen, not a lie detector.
1 in the Claude listing