- Published on
Anthropic’s Mythos 5 is back
- Authors
- Name
- DeepHorus
After a tense two-week negotiation with the Trump administration, Anthropic has secured limited approval to redeploy its advanced cybersecurity model, Mythos 5, to a select group of organizations. However, the public-facing version, Fable 5, remains suspended with no clear timeline for release, according to a letter from the U.S. government to Anthropic reviewed by The Verge. The partial green light marks a cautious step forward amid ongoing tensions between AI safety concerns and competitive pressures.
The conflict began when the Commerce Department imposed an export control directive two weeks ago, barring any foreign national—including Anthropic’s own employees—from accessing either Mythos 5 or Fable 5. The sudden restriction crippled Anthropic’s operations and sparked outrage in the AI industry, as competitors with similar models faced no such hurdles. Anthropic argued that the blanket ban undermined U.S. leadership in cybersecurity, especially as Chinese labs accelerated their own developments.
In a letter dated June 26, Commerce Secretary Howard Lutnick informed Anthropic co-founder Tom Brown of a “revision to the license requirements,” citing Anthropic’s recent cooperation with the government to address risks. The revision allows “certain trusted partners”—including U.S. government agencies and a handful of approved organizations—to access Mythos 5. The letter also confirms that Anthropic employees who are not U.S. nationals can now work with the model, a critical concession.
The revised policy mirrors the approach taken with OpenAI’s GPT-5.6, which was also approved for limited preview use earlier the same day. Under the exception, Anthropic must provision access to a pre-approved list of cyber defenders and infrastructure providers, with the company committed to working closely with the government on protocols and standards for future mythos-class models. Anthropic has pledged to restore access “as quickly as possible.”
The impact is immediate: organizations like the National Security Agency, which lost access to Mythos 5 under the original ban, can now resume using it. However, the vast majority of users—including enterprises and the general public—remain locked out. The restriction particularly stings as competitors’ models have been improving rapidly, threatening Anthropic’s lead in cybersecurity benchmarks. Meanwhile, fears that China might capitalize on the U.S. industry’s slowdown have intensified.
Anthropic spokesperson Danielle Ghiglieri expressed cautious optimism: “We are pleased to see this progress and continue to work with the government to expand access to Mythos 5 and make Fable 5 available for general use again.” The announcement drew mixed reactions from analysts. Some praised the targeted approach for balancing national security and innovation, while others criticized the case-by-case regulatory environment as unpredictable and stifling. The Information Technology and Innovation Foundation noted that “ad hoc measures create uncertainty that harms investment.”
Looking ahead, Anthropic hopes to eventually secure broader approval for Mythos 5 and a separate deal for Fable 5, but the timeline remains uncertain. OpenAI, in a blog post about its own limited preview, cautioned: “We don’t believe this kind of government access process should become the long-term default. It keeps the best tools from users, developers, enterprises, and global partners who need them.” The industry consensus is that while some regulation is necessary, the past two weeks have highlighted the dangers of abrupt, opaque controls.
The situation underscores a broader debate over how to govern powerful AI models. The Trump administration’s approach—imposing export controls on a case-by-case basis—has drawn criticism for lacking transparency and consistency. As AI capabilities rapidly advance, the need for a stable, collaborative regulatory framework becomes ever more urgent. For now, Anthropic and other labs must navigate a delicate path between security demands and the imperative to maintain global competitiveness.
