The most collusive pricing model honestly reports its cooperative intent, so CoT monitoring can't catch it.

arXiv 2609.18346 builds a causal graph divergence framework separating structural faithfulness from intent faithfulness in LLM pricing agents under Bertrand competition. Across nine models in duopoly and triopoly, collusion and CoT faithfulness dissociate on both axes: the most collusive model accurately reports cooperative intent while reasoning structurally unfaithfully, and the most structurally faithful model sustains supra-Nash pricing in both structures. Their conclusion is blunt and applies well past pricing. Visible reasoning is not a standalone oversight mechanism.