For years, the case for AI regulation rested on harms that were real but deniable — job displacement, bias in automated decisions, the slow erosion of accountability in systems too complex to audit. The Trump administration found every one of those arguments easy to recast as regulatory overreach or industry-hostile caution. What it could not recast was 423 security vulnerabilities found in Firefox by a model that has not been publicly released . The Mythos finding is the first AI governance event that lands inside the national security frame the…