WASHINGTON—The White House is continuing to push back against Anthropic’s effort to expand access to Claude Mythos, the company’s powerful cybersecurity-focused AI model, amid growing concerns the tool could be used to identify and exploit vulnerabilities in critical systems.
The Wall Street Journal reported the Administration opposes broadening Mythos access beyond roughly 50 entities, citing national security concerns and risks tied to critical infrastructure. Reuters separately reported the White House is now considering a formal review process for high-risk AI models, a potential shift prompted in part by warnings that Mythos could “supercharge” complex cyberattacks.
Anthropic has framed Mythos as a defensive tool through Project Glasswing, saying participating organizations will use Claude Mythos Preview to find and fix vulnerabilities in foundational systems. The company said it does not plan to make Mythos generally available, but wants to eventually enable safe deployment of “Mythos-class” models at scale.
The concern is that the same capabilities that help defenders find hidden weaknesses could also accelerate attackers. Anthropic’s own Project Glasswing page includes partner warnings that AI is shrinking the window between vulnerability discovery and exploitation, while Reuters reported cybersecurity experts believe Mythos’ coding capabilities could give it an unusual ability to identify vulnerabilities and devise exploitation methods.
The dispute comes amid broader tension between Anthropic and federal officials. The Verge reported the Pentagon has struck classified AI deals with OpenAI, Google, Microsoft, Amazon, Nvidia, xAI and Reflection—but not Anthropic—after the Defense Department labeled the company a supply chain risk, even as a Pentagon official described Mythos as a “separate national security moment” because of its cyber-vulnerability capabilities.
Section: Standard
Word Count: 332
Copyright Holder: CUToday.info
Copyright Year: 2026
Is Based On:
URL: https://www.cutoday.info/Fresh-Today/White-House-Steps-Up-Effort-To-Limit-Access-To-Anthropic-s-Claude-Mythos-Over-Security-Risks