Scoop: Top U.S. Cyber Agency Doesn't Have Access to Anthropic's Powerful Hacking Model

Why It Matters

Without Mythos, CISA’s ability to identify AI‑driven threats and share actionable intelligence with banks, power plants, and other critical sectors is severely weakened, widening the nation’s cyber‑defense gap.

Key Takeaways

• Anthropic's Mythos Preview can rapidly discover and exploit vulnerabilities
• Over 40 firms, including NSA, are testing Mythos; CISA is excluded
• CISA faces budget cuts up to $707 million and workforce reductions
• Trump administration shifted cyber policy control to White House cyber director
• Critical infrastructure loses a key threat‑intelligence partner amid AI risks

Pulse Analysis

The rollout of Anthropic’s Mythos Preview underscores a new frontier where generative AI can be weaponized to uncover software bugs at unprecedented speed. While the model promises to harden defenses by exposing hidden flaws, its dual‑use nature forces companies to restrict access, creating a selective ecosystem of early adopters. Organizations that have secured Mythos are already running internal red‑team exercises, effectively turning the AI into a virtual penetration tester that can outpace traditional tools.

CISA’s exclusion from this testing pool is especially concerning given its historic role as the nation’s central hub for cyber threat intelligence. The agency’s dwindling budget—potentially slashed by $707 million—and a workforce reduced by more than a third limit its capacity to procure cutting‑edge technologies. As the Trump administration reassigns cyber‑policy influence to the White House’s national cyber director, CISA’s diminished resources could erode the coordinated response mechanisms that critical infrastructure sectors rely on during large‑scale attacks.

Stakeholders should monitor how the White House negotiates broader civilian agency access to Mythos, a process that may set precedents for AI governance in national security. If CISA remains sidelined, private sector entities may need to fill the intelligence vacuum, potentially leading to fragmented defense postures. Conversely, a successful integration of Mythos across government could accelerate vulnerability remediation, but only if the agency can secure the necessary funding and talent to operationalize the technology effectively.