A source familiar with the situation, who requested anonymity due to the sensitive nature of the matter, indicated that more than 100 companies and institutions, including many Fortune 500 firms, will now be able to access Mythos 5.
Since June 12, we’ve been working closely with the US government to restore access to Claude Mythos 5 and Fable 5. Today, the government notified us that Mythos 5, our strongest cybersecurity model, can be redeployed to a set of US organizations that operate and defend critical…
— Anthropic (@AnthropicAI) June 27, 2026
The concern over the potential misuse of advanced AI systems by military intelligence in nations like China and Russia has led the Trump administration to adopt a stringent oversight policy regarding the release of models from Anthropic and its competitor, OpenAI.
Earlier in the day, OpenAI announced it would postpone the public launch of GPT-5.6 at the US government’s request, restricting access to a select group of vetted partners whose identities were communicated to the authorities.
Following the government’s export control order on June 12, Anthropic had abruptly disabled access to its most advanced AI models—Mythos 5 and Fable 5—for all users.
“The government informed us today that Mythos 5, our most powerful cybersecurity model, can now be redeployed to US organizations that manage and protect critical infrastructure,” Anthropic stated on Friday, June 26.
“We’re expediting access for these organizations and are actively collaborating with the government to broaden access to Mythos 5 and to reintroduce Fable 5 for general use,” it added.
GOVERNMENT CRITICIZED FOR SELECTIVE ACCESS
The government’s vetting process for companies eligible to access Mythos has faced considerable backlash.
“It’s unclear how these companies are chosen and why others are excluded,” remarked John Coleman, legislative counsel for the nonpartisan free-speech nonprofit, the Foundation for Individual Rights and Expression.
“This grants excessive power to the government. There’s a lack of transparency, which raises legal concerns.”
Sam Altman, CEO of OpenAI, echoed apprehensions regarding the government’s role in selecting access to top models in a post on X.
“Extensive safety testing is a wise approach, but I’m uneasy about the government deciding who gets to use these technologies,” he noted.
Experts warn that in the wrong hands, Mythos models could significantly accelerate sophisticated cyberattacks, particularly in sectors like banking that depend on intricate, interlinked, and often outdated technological frameworks.
A letter from US Commerce Secretary Howard Lutnick to Anthropic noted “significant progress” in their collaboration with the government to mitigate “risks associated with the Covered Models.”
The specific safeguards implemented remain unclear. Earlier this month, Anthropic indicated that the government suspected a method to circumvent a safeguard preventing Fable 5 from identifying software vulnerabilities.
NO EXPORT LICENSE REQUIRED FOR NON-US CITIZENS
Lutnick’s letter stated that an export license will no longer be necessary for Mythos 5 when shared with trusted companies and their employees who are not US citizens, as well as with Anthropic’s non-US citizen staff, although licensing restrictions will persist for companies not on the approved list.
The source mentioned that many approved companies are participants in Anthropic’s Project Glasswing, which includes around 100 well-known tech firms and institutions.
Additionally, the government is progressing towards allowing Anthropic to launch Fable soon, although a timeline is still uncertain, according to the source.
Both Fable 5 and Mythos utilize the same foundational AI model; however, Fable 5 is intended for broader public access, while more safeguards are in place for Mythos.
Both Anthropic and OpenAI are preparing for public offerings.
Anthropic’s relationship with the US government has been notably turbulent. The company declined to allow the military to use its AI models for domestic surveillance and fully autonomous weaponry, leading to its placement on a national security blacklist.
The government’s restrictions on both Anthropic and OpenAI follow Trump’s executive order this month, establishing a voluntary framework for AI developers to present “covered frontier models” to the US government for up to 30 days before releasing them to selected partners.
The administration’s latest directive is seen as “a practical interim measure but leaves unresolved the broader issue of how companies can widely release updated models,” noted Kate Koren, an analyst at the Center for Strategic and International Studies in Washington and a former Commerce Department official.
“The absence of a system that allows US companies to freely release new models increases the risk that China will catch up,” she added.
