AnthropicA company that spent $8 billion positioningon itselfAI assafety theand mostlectured safety-focusedCongress AIabout labexistential onrisk earth,got then exposed itsby entire codebase through a config file any juniormid-level engineer would've caughtcatch. The same companylab warningthe CongressPentagon aboutflagged existentialas AIa supply chain risk couldn'taccidentally handlepublished a source map to npm — a basic codemistake review.that Thisunmasked isn'tClaude Mythos, a minormodel slipso —dangerous Anthropic itself admits it's acould fundamentaloutpace credibilityhuman collapsedefenders.