When Governance Has No Protocol: Solving the DoD/Anthropic Standoff with Negotiated Solvency

When two institutions collide over an AI system’s operational constraints, the standard diagnosis is a values conflict: one side wants capability, the other wants safety. The debate proceeds as if the solution is finding the right point on that axis—more access, fewer guardrails, or vice versa. That diagnosis misses the failure mode. The failure is not values. It is the absence of a shared protocol for negotiating compliance under live operational pressure. A live instance of this failure mode was reported in February 2026, when Reuters described an escalating DoD–Anthropic dispute over military safeguards for Claude and the possibility of Defense Production Act action. The argument in this paper does not depend on any particular detail of that report. It depends on the structural pattern: two sovereign parties, each acting within mandate, reaching an impasse because neither has protocol-native means to negotiate deviation and oversight. This is not a policy failure. It is a stream integrity failure.