The US government has officially cleared Anthropic to redeploy Claude Mythos 5 to critical-infrastructure defenders, ending a sudden two-week export-control freeze.
The decision allows the company to restore access to Claude Mythos 5 for a select group of US organisations. These entities operate in vital sectors like energy, financial services, and telecommunications. The move partially lifts a strict export-control freeze that had taken the company’s two most capable artificial intelligence systems offline since mid-June.
What the Claude Mythos 5 Clearance Means
The clearance came in an official letter from US Commerce Secretary Howard Lutnick to Anthropic. The authorization covers a defined set of vetted US entities, reported to number more than 100 institutions, rather than a full return to public availability. The targeted organisations align with the critical-infrastructure sectors defined under the US Cybersecurity and Infrastructure Security Agency (CISA) framework. These sectors increasingly lean on advanced tools to triage vulnerabilities and accelerate incident response.
The Origin of the Export Freeze
The redeployment partly reverses a directive issued two weeks earlier. On June 12, the Commerce Department’s Bureau of Industry and Security invoked the Export Administration Regulations. This framework governs dual-use technology with national-security implications. It barred access to the models by any foreign national, whether inside or outside the United States, including Anthropic’s own foreign-national staff.
Because Anthropic does not verify user nationality in real time, the company had to disable both models globally for all customers to remain compliant. Access to its other models, including Opus 4.8, was unaffected. The government cited national-security concerns tied to a reported method of “jailbreaking” Fable 5, which is the public-facing version of the model.
Project Glasswing and Model Capabilities
Previously, Claude Mythos 5 was only available through Project Glasswing, a restricted programme run with the US government. Its partners reportedly included infrastructure providers such as Cisco and banks such as JPMorgan Chase. While Mythos 5 and Fable 5 share the same underlying model, Fable 5 ships with stronger guardrails that block responses in high-risk areas such as cybersecurity and biology.
Early testing showed the technology was particularly effective at autonomously discovering software vulnerabilities. This included finding long-standing flaws in widely used open-source software. Fable 5 general use remains suspended pending further discussions.
Strategic Implications for Saudi and GCC Enterprises
This episode is a notable signal for ICT buyers across Saudi Arabia and the wider GCC. The original directive’s reach over foreign nationals meant non-US users and organisations were swept up in the suspension. This serves as a reminder that access to frontier AI from US vendors can now hinge on national-security policy rather than commercial terms alone.
As governments treat the most capable models as controlled strategic assets, regional enterprises building production workflows on a single frontier provider face a new category of continuity risk. This strengthens the case for vendor diversification and contingency planning across the regional economy and telecommunications sectors.
