Frontier AI Is Now Critical Infrastructure

The report describes the U.S. government’s emergency export-control-driven shutdown of Anthropic’s Claude Fable 5 and Mythos 5 models after red-team testing and guardrail bypasses demonstrated the models could rapidly map vulnerabilities within classified systems; it outlines the policy rationale, alleged supply-chain disclosures from Amazon and the NSA, potential restricted re-release timelines, and significant industry pushback arguing the blackout harms defensive capabilities and increases reliance on open-weight or local models.