Anthropic's Claude Mythos 5 Cleared for Redeployment to US Critical Infrastructure After Two-Week Government Suspension

In one of the most consequential episodes in the short history of frontier AI governance, Anthropic has secured partial clearance from the United States government to restore access to Claude Mythos 5, its most powerful cybersecurity model. According to Anthropic's official statement posted on June 27, 2026, the US government notified the company that Mythos 5 can be redeployed to a defined set of US organisations that operate and defend critical infrastructure. The model had been suspended on June 12, just three days after its launch via Project Glasswing, following a government export control directive citing national security concerns. US Secretary of Commerce Howard Lutnick signed off on the restoration, writing that he had determined appropriate safeguards were in place, as reported by CNBC and Semafor.

Claude Mythos 5 is no ordinary AI model. Anthropic publicly described it in April 2026 as a potential 'cybersecurity reckoning' — a system so capable that the company initially chose not to release it to the public at all. According to Anthropic and cybersecurity news site GBHackers, the model demonstrated an unprecedented ability to autonomously discover software vulnerabilities, uncovering thousands of high-severity flaws across every major operating system and web browser, including a 27-year-old vulnerability in OpenBSD and a 16-year-old flaw in FFmpeg. The companion model Claude Fable 5, a version of the same underlying system with additional safety classifiers, was launched for enterprise use on June 9 but was also caught in the shutdown, as Anthropic was unable to technically comply with the government's directive without disabling both models entirely.

The restoration is notably partial and structured. Commerce Secretary Lutnick's authorisation covered roughly 100 pre-approved trusted partners, including selected federal agencies and critical infrastructure providers, according to reporting by Digit.in. Anthropic confirmed it is continuing to work with federal authorities toward a broader expansion of Mythos 5 access and the eventual return of Fable 5 for general enterprise use, suggesting a phased clearance process that mirrors protocols more commonly associated with sensitive military and defence technologies. Crucially, the broader dispute between Anthropic and the Department of Defense — which began in February 2026 when the DOD demanded access for autonomous weapons applications — remains on a separate and still-active legal track.

The episode crystallises a tension that is now structuring the entire frontier AI industry: the US government increasingly treats the most capable AI models as dual-use technologies subject to export control regimes, not as commercial software. As reporting from Build Fast With AI noted, the Commerce Department's framework explicitly requires export licences for non-allied entities seeking access to Mythos 5, mirroring the treatment of advanced semiconductors. The two-week negotiation was reportedly complicated by interpersonal friction at the highest levels, with White House insiders, as reported by iTech Post, characterising communications with Anthropic's leadership as difficult before talks improved. OpenAI, watching the confrontation from the sidelines, publicly stated it believes in broad access while opposing making per-customer government approval a permanent norm.

For the cybersecurity community, the stakes of this precedent are immense. Defenders operating within critical sectors including energy, healthcare, financial services, and telecommunications had been left without access to the most capable AI-driven vulnerability detection tools during the review window, a gap that adversaries do not face by definition. As Anthropic works through the phased authorisation to restore Fable 5 for broader enterprise use, the industry will watch closely to see whether Washington ultimately adopts a clearance model that permanently tiers access to frontier AI by national security criteria — and how allies and competitors in Europe and Asia respond to being structurally excluded from the most powerful defensive tools in the arsenal.