Washington Allows Anthropic to Release Mythos 5 Model to Trusted U.S. Institutions; Fable 5 Expected to Be Unlocked Early Next Week
The artificial intelligence company Anthropic said on Friday that the U.S. government has permitted it to release its powerful Claude Mythos 5 AI model to a select group of "trusted" American institutions; this move partially reverses a two-week-old pause on access issued due to national security concerns. A source familiar with the matter revealed that over 100 companies and organizations will now be able to use Mythos 5, including many Fortune 500 firms.
Concerned about the potential misuse of advanced AI systems by military-intelligence agencies in China, Russia, or other nations of concern, the Trump administration has taken a hardline approach, regulating Anthropic and its competitor OpenAI’s release of cutting-edge AI models.
Earlier that day, OpenAI stated that, at the request of the U.S. government, it would delay the full public launch of GPT-5.6 and restrict access to a small number of vetted partner organizations, while also notifying relevant authorities about these partners.
After the Trump administration issued an export control order on June 12, Anthropic had abruptly disabled its most advanced AI models—Mythos 5 and Fable 5—for all users.
In its statement on Friday, Anthropic said: “Today, this administration has informed us that Mythos 5—the most powerful cybersecurity model we have—can be redeployed to a select group of U.S. institutions responsible for operating and protecting critical infrastructure.”
Anthropic added: “We are rapidly restoring access for these institutions and continuing to work with this administration to expand the use of Mythos 5 and bring Fable 5 back to public availability.”
The Trump administration’s review of which companies were granted access to Mythos has drawn widespread criticism.
OpenAI CEO Sam Altman posted on X expressing concern over the government’s decision-making regarding who can use top-tier models. He wrote: “Conducting broad security testing isn’t bad—but I just dislike the idea of the government picking customers.”
Experts warn that if misused, Mythos models could significantly accelerate complex cyberattacks, particularly in sectors like banking that rely on intricate, interconnected, and often decades-old technological systems.
U.S. Commerce Secretary Howard Lutnick previously wrote to Anthropic, stating the company had made “significant progress” in collaborating with the Trump administration to address risks associated with “Covered Models.”
It remains unclear exactly what safeguards have been implemented. Earlier this month, Anthropic said it understood the Trump administration’s concern that there may exist some method—referred to as a “jailbreak”—to bypass protective measures, potentially enabling the Fable 5 model to be used for identifying software vulnerabilities.
In his letter, Lutnick stated that no export license would be required when providing Mythos 5 to trusted companies and their non-U.S. citizen employees, or to Anthropic’s own non-U.S. citizen staff. However, licensing restrictions would still apply to companies not on the approved list.
Sources say many of the approved companies are part of Anthropic’s “Project Glasswing,” which includes around 100 well-known tech companies and institutions.
Fable 5 and Mythos are built on the same underlying AI model, but Fable 5 is designed for broad public release, whereas Mythos has had certain safety restrictions lifted. Both Anthropic and OpenAI plan to go public.
According to Axios on Saturday, citing sources familiar with the matter, the Trump administration is nearing a decision to allow Anthropic to restore access to its powerful Fable 5 model—previously taken offline for 15 days due to government security concerns.
Insiders expect the restrictions on the Fable 5 model could be lifted as early as next week. Another source said related discussions are expected to continue over the weekend, and Anthropic anticipates regaining access to the Fable model very soon.
The shutdown of the Fable 5 model has been unprecedented and deeply impactful, especially for developers and early users without technical backgrounds—a top-tier model already in users’ hands was suddenly pulled offline due to intervention by the Trump administration. Progress in securing approval for Fable 5 marks a sign of easing in the four-month standoff between the Trump administration and Anthropic.
Source: rfi
Original article: toutiao.com/article/1869200701115404/
Disclaimer: This article represents the personal views of the author