Anthropic's Claude Mythos 5 Gets Green Light for Over 100 US Institutions — Is Fable 5 Next?

The United States government has officially removed its export restrictions on Anthropic's Claude Mythos 5, clearing the advanced AI model for deployment across more than 100 American institutions, including major corporations and federal agencies. The announcement came on Friday and marked a significant turning point in a two-week standoff between the Trump administration and the AI company.

The decision specifically benefits Mythos 5 — the enterprise-focused variant — while the consumer-facing Fable 5 remains unavailable to the general public for the time being.

**Commerce Department Grants Access to Trusted Partners**

Commerce Secretary Howard Lutnick formalized the resolution through a letter addressed to Anthropic's compute chief, Tom Brown. According to the letter, organizations listed in Annex A are no longer required to obtain a special export license to access Claude Mythos 5. Lutnick stated that "appropriate safeguards are in place to permit certain trusted partners" to utilize the model.

Senior Anthropic executives had traveled to Washington, D.C., to hold direct meetings with administration officials during the dispute, according to CNBC. Their efforts appear to have contributed to the favorable outcome.

Prior to the export block, Mythos 5 was operating within Project Glasswing — a cybersecurity vulnerability-hunting initiative involving approximately 150 organizations across more than 15 countries. The model had demonstrated remarkable capabilities, reportedly identifying critical flaws in classified government systems within hours of initial testing.

The restrictions had been triggered in part by concerns raised by Amazon, one of Anthropic's largest investors. Amazon researchers warned that Fable 5, the publicly accessible version, could potentially be manipulated through jailbreaking techniques for malicious purposes.

**Fable 5's Future Remains Uncertain**

Sources familiar with ongoing negotiations indicate that a path forward for Fable 5 is taking shape, though no firm release date has been confirmed. Before the block was imposed, Fable 5 was available to any subscriber and had briefly earned recognition as the most capable AI tool accessible to the general public.

This episode is solidifying into a broader regulatory framework for frontier AI. A June 2 executive order established a voluntary review channel allowing developers to submit models for federal cybersecurity evaluation up to 30 days before launch. This extends Washington's existing AI governance approach — which has focused heavily on restricting chip exports to China — into direct oversight of model access and distribution.

Anthropically, OpenAI followed a parallel path on the same Friday, restricting its most powerful GPT-5.6 tier, known as Sol, to roughly 20 government-approved partners, while making the lighter Terra and Luna versions publicly available.

**Security Community Pushes Back**

The original restrictions were partly fueled by concerns over potential Chinese access to the technology, linked to SK Telecom — a South Korean carrier that had joined Project Glasswing in early June before losing its access. SK Telecom has publicly denied any connections to China.

Dozens of cybersecurity leaders signed an open letter urging the administration to lift the controls. The letter, organized by former Facebook security chief Alex Stamos, gathered signatures from prominent technology firms including Nvidia, Adobe, and Zoom.

International frustration has also been growing, with European and other allied nations finding themselves suddenly reliant on Washington's approval for access to cutting-edge AI tools. Whether Fable 5 will receive the same clearance as Mythos 5 may become apparent within the coming days, as negotiations continue behind the scenes.