Brijesh Desai 
CISA Gets Full Access to Anthropic’s Mythos Preview AI model, received access around June 9, 2026. After being locked out since April, White House hasn’t set clear usage parameters yet. Previously only NSA, Commerce had access.
The Cybersecurity and Infrastructure Security Agency (CISA) now has full access to Anthropic’s flagship Mythos Preview model, according to a U.S. official and a second person familiar with the matter. CISA received access around June 9, 2026 (approximately a week ago), marking a significant reversal from April 2026 when the agency was locked out of the AI security tool it arguably needs most.
However, the White House has not yet set clear parameters for how CISA should use the model, leaving the agency without formal guidance on its usage.
The Timeline: From Locked Out to Full Access
The turnaround happened in just 6 days from “imminent” (June 10) to “received access” (June 9).
What Mythos Preview Is: The “Cyber-Weapon” AI
Mythos Capabilities:
The model can rapidly spot cybersecurity vulnerabilities and offer the ability to exploit them, making it potentially dangerous if misused.
Why Access Is Restricted:
- Offensive capabilities — Can identify and exploit flaws
- Power balance concern — Could upset cybersecurity power balance
- Dangerous potential — Too powerful for broad public release
- Supply-chain risk — Pentagon labeled Anthropic a “supply-chain risk”
Who Had Access Before CISA
Initial Mythos Preview Recipients (April 2026):
The Irony:
CISA, America’s lead cybersecurity agency and the nerve center for national cybersecurity coordination, didn’t have access while agencies with more specialized national security mandates (like NSA) were already testing it.
Why CISA Was Locked Out Initially
The Pentagon-Anthropic Conflict:
The Pentagon blacklisted Anthropic, but the NSA—part of the Department of Defense—was secretly using Mythos despite the ban.
CISA’s Role vs. NSA’s Role:
CISA’s mission is to protect critical tech infrastructure and coordinate cybersecurity across the nation, making Mythos arguably more critical for CISA than NSA.
What CISA Will Use Mythos For
Proposed Use Case (from White House discussions):
The White House is weighing having CISA leverage Mythos to scan federal agencies’ networks.
Current Status: Full Access, No Parameters
What CISA Has Now:
What’s Still Unclear:
-
How CISA should use Mythos (no formal guidance)
-
What restrictions apply to CISA’s usage
-
Timeline for White House parameters
The agency has access but operates without clear usage parameters.
The White House’s Role in Mythos Access
Ongoing Government Access Plans:
Gregory Barbaccia (White House OMB Chief Information Officer) said they’re working closely with model providers, industry partners, and intelligence community to ensure appropriate guardrails before releasing modified version to agencies.
The Mythos Paradox: CISA Locked Out While Unauthorized Users Had Access
The Contradiction:
What Happened:
- A small group of individuals gained unauthorized access to Claude Mythos through a third-party vendor environment
- The group already had permission to view Anthropic’s AI systems due to previous contractor work
- They were using Mythos since gaining access, but not for malicious purposes (aiming to avoid detection)
- Anthropic is investigating the unauthorized access claim
The irony: random unauthorized users got access before America’s primary cybersecurity agency.
Why This Matters: National Cybersecurity Implications
CISA’s Mission:
- Protect critical tech infrastructure in the US
- Central coordinator for national cybersecurity
- Infrastructure protection across federal agencies
Mythos’s Strategic Value:
- Autonomous vulnerability discovery at scale
- Can find thousands of critical flaws across software systems
- Offensive cybersecurity capabilities — can exploit vulnerabilities
- Defensive tool for infrastructure protection
The Power Balance:
Mythos could upset the cybersecurity power balance by giving whoever has it unprecedented ability to find and exploit flaws.
The Bottom Line
CISA now has full access to Anthropic’s Mythos Preview model, receiving access around June 9, 2026 after being locked out since April 2026. This marks a significant reversal from April 21 when Axios reported CISA lacked access while NSA and Commerce Department were already testing Mythos. The White House has not yet set clear parameters for how CISA should use the model, leaving the agency without formal usage guidance. Mythos is a purpose-built autonomous vulnerability-discovery AI capable of finding thousands of critical flaws across major software systems, making it a powerful “cyber-weapon” restricted to ~40 vetted organizations under Project Glasswing. The Pentagon had labeled Anthropic a “supply-chain risk” and barred the company from defense contracts, but NSA (part of DoD) secretly used Mythos despite the ban. Ironically, a random Discord group got unauthorized access before CISA did. CISA will likely use Mythos to scan federal agencies’ networks for public-facing vulnerabilities and security flaws.
Quick Summary
CISA now has full Mythos Preview access, received access around June 9, 2026 (one week ago). After being locked out since April 21, 2026 when Axios reported CISA lacked access while NSA and Commerce Dept already testing Mythos. White House hasn’t set clear usage parameters yet. Mythos: autonomous vulnerability-discovery AI finding thousands of critical flaws across software systems, “cyber-weapon” restricted to ~40 organizations under Project Glasswing. Pentagon labeled Anthropic “supply-chain risk,” barred from defense contracts, but NSA secretly used Mythos despite ban. CISA will scan federal agencies’ networks for public-facing vulnerabilities. Random Discord group got unauthorized access before CISA did. CISA: America’s lead cybersecurity agency, nerve center for national cybersecurity coordination, protecting critical tech infrastructure.