Anthropic's Mythos 5, a flagship cybersecurity AI model, has been restored to limited availability after a two-week standoff with the Trump administration. The Commerce Department, through a letter sent by Secretary Howard Lutnick on June 26, authorized a small group of trusted organizations—including key cyber defense agencies and infrastructure providers—to access the model once again. The decision marks a partial resolution to a broader regulatory conflict that had paralyzed Anthropic's operations and drawn sharp criticism from industry leaders and national security officials.
The letter, addressed to Anthropic co-founder Tom Brown, cited a recent collaboration between the company and the US government to address safety concerns associated with Mythos 5 and its consumer-grade counterpart, Fable 5. Under the revised license requirements, Anthropic employees who are not US nationals and members of approved organizations can now use Mythos 5. The government, however, stopped short of lifting the broader export control directive issued two weeks ago, which had barred any foreign national—including Anthropic's own staff—from accessing either model.
Anthropic spokesperson Danielle Ghiglieri confirmed the development in a statement, noting that the company had received official notice from the US government and was working to quickly restore access for the approved providers. She expressed hope that this progress would lead to expanded access to Mythos 5 and a full return of Fable 5 for general use. The public-facing model remains unavailable, and no timeline has been provided for its rollout.
The Deal in Context
The partial restoration of Mythos 5 mirrors a similar arrangement that OpenAI secured for its GPT-5.6 model, which was announced earlier on the same day. In both cases, the Trump administration insisted on a limited preview model, restricting access to trusted enterprises and US government agencies. This case-by-case regulatory approach has drawn criticism from tech leaders who argue it stifles innovation and puts American companies at a disadvantage against competitors in China and elsewhere.
Pressure on the administration had been building for weeks. National security agencies, including the National Security Agency (NSA), lost access to Mythos 5 during the standoff, raising alarms about the US's ability to defend against cyber threats. At the same time, rival labs were making rapid progress on their own cybersecurity-focused models, pulling ahead on key benchmarks and securing government contracts. The situation created a paradoxical environment where the US government's own defenses were weakened while it tried to regulate its leading AI developers.
OpenAI, in its blog post announcing GPT-5.6, made no secret of its displeasure with the process. The company wrote that it did not believe this kind of government access process should become the long-term default, arguing that it kept the best tools from users, developers, and global partners. OpenAI framed the short-term step as a necessary compromise to achieve broader availability in the coming weeks, while working with the administration to develop a more sustainable framework for future model releases.
Background: The Rise of Anthropic and Mythos
Anthropic was founded in 2021 by former OpenAI employees, including Tom Brown, who sought to build AI systems that are safer and more aligned with human interests. The company has positioned itself as a leader in AI safety research, focusing on constitutional AI and responsible deployment. Mythos 5, released earlier in 2026, represents the culmination of years of development in cybersecurity applications, capable of detecting and neutralizing threats that traditional tools miss.
Fable 5, the consumer-facing version of the Mythos-class model, was intended to bring similar capabilities to a broader audience, including small businesses and individual developers. Its indefinite delay has frustrated many in the tech community who see it as a critical tool for improving security across the internet.
The current regulatory battle can be traced back to the Trump administration's Executive Order on AI, signed in early 2025, which empowered the Commerce Department to impose export controls on advanced AI models deemed critical to national security. Anthropic's models were among the first to be caught in this net, alongside offerings from OpenAI and other labs. The process has been criticized as opaque and inconsistent, with companies forced to negotiate case by case without clear guidelines.
The Broader Impact
The partial unblocking of Mythos 5 is likely to be seen as a temporary reprieve rather than a lasting solution. The letter from Secretary Lutnick explicitly stated that all other requirements from the June 12 directive remain in effect until further notice, and that he reserved the right to reevaluate the scope of license requirements should circumstances change. This means that both Anthropic and its approved partners remain in a fragile state, subject to potential new restrictions at any time.
Industry observers note that the approach risks creating a two-tier system where only a handful of well-connected organizations have access to the most powerful AI tools. This could exacerbate existing inequalities in cybersecurity, leaving smaller companies and nonprofits vulnerable to attacks that larger players could thwart. Furthermore, the lack of a clear process for obtaining approval may deter foreign entities from partnering with American firms, undermining the US's leadership in the global AI ecosystem.
On the other hand, proponents of the administration's stance argue that strong oversight is necessary to prevent advanced AI from being used maliciously or falling into the hands of adversaries. They point to China's rapid AI development as a reason for caution, though critics counter that restricting American companies only hands the advantage to foreign rivals.
The coming weeks will be critical as Anthropic continues to negotiate for full access to Mythos 5 and the release of Fable 5. Meanwhile, the AI industry as a whole watches closely, hoping that the regulatory framework will evolve into something more predictable and sustainable. For now, the restoration of Mythos 5 for a select few offers a glimmer of progress, but the underlying tensions remain unresolved.
Anthropic's immediate priority is to provision the approved set of providers—believed to include major cloud operators, government cybersecurity centers, and critical infrastructure operators—and restore their access as quickly as possible. The company has also committed to working with the US government on protocols and standards for future releases of Mythos-class models, in an effort to build trust and avoid further disruptions.
Source: The Verge News