Warner's critical-infrastructure bill names an AI model as the threat

Senate Intelligence ranking member Mark Warner has introduced the Combat Emerging Threats to Critical Infrastructure Act, per Inside Cybersecurity, directing CISA to take the lead on updating the sector-specific plans called for under the Biden-era 2024 national security memorandum on critical infrastructure. The justification is unusually specific about what changed: the release announcement cites “the development of Anthropic’s Claude Mythos — an AI model capable of identifying and exploiting vulnerabilities in our country’s cybersecurity infrastructure” as the reason the plans need refreshing.

Communications is one of those sector-specific plans, which is the hook here. A bill that re-opens the sector frameworks is a bill that eventually touches signaling, numbering, and call-authentication infrastructure as critical-infrastructure questions rather than purely FCC ones — a different regulatory venue for the same trust problems. And the framing is worth noting on its own: a sitting senator is now writing a named frontier model into legislative findings as a live offensive cyber threat, not a hypothetical.

The Mythos reference is the line to track. It connects this directly to the export-control fight playing out over Anthropic’s models — the same capability that has the government invoking national-security authorities to restrict access is now being cited on the Hill as grounds to harden the sectors. When AI-as-cyber-weapon shows up in a critical-infrastructure bill, the telecom trust beat and the AI-governance beat have started to converge.