US Commerce Department Permits Anthropic to Release Mythos 5 AI Model to Vetted Domestic Organizations

The US Commerce Department on June 26 notified Anthropic that its Mythos 5 artificial intelligence model may be distributed to a set of "trusted partners" — more than 100 companies, Fortune 500 firms, and federal agencies — without an export license, partially reversing restrictions imposed two weeks earlier [2] [3]. Commerce Secretary Howard Lutnick stated in a letter that "Anthropic has worked with the US government to address risks associated with the Covered Models" and that the efforts had "yielded significant progress"

. A Commerce Department spokesperson said the department had "worked diligently to ensure America remains the global leader in AI while safeguarding our security"

[2] [2]. Anthropic's consumer-facing model, Fable 5, remains restricted, though an unnamed source familiar with the directive said the government is "moving towards allowing release of Fable as well," without a clear timeline

[5]

[2][.](#src-003)

[3]Anthropic described Mythos 5 as "our strongest cybersecurity model" and said it welcomed its redeployment "to a set of US organisations that operate and defend critical infrastructure" [3]. Many of the approved organizations belong to Anthropic's Project Glasswing, a consortium of roughly 100 technology companies and institutions

. CNBC confirmed the approximate number of trusted partners and noted that similar limitations had been applied to rival OpenAI's latest models

[3]. A report by Euronews's Russian-language service stated that Mythos had identified vulnerabilities in classified US systems within hours during testing, a finding cited in coverage of the period leading up to the restrictions

[9]. [19]The backstory of the restrictions centers on a disputed sequence of events. David Sacks, co-lead of the Trump administration's council of technology and science advisers, said that Anthropic CEO Dario Amodei "came to Washington a few months ago, back in April, and basically said that he had created a cyber weapon called Mythos" [4]. Sacks further stated that a trusted partner had discovered a jailbreak method capable of bypassing the safety guardrails separating Fable 5 from Mythos's unrestricted cybersecurity capabilities, and that Amodei refused the government's request to patch or withdraw Fable 5, prompting the export-control order

. Anthropic has mounted a legal challenge to its designation as a national security and supply chain risk by the Pentagon, a designation that Defense Secretary Pete Hegseth declined to reconsider

[5]. [8]The process by which organizations are approved or excluded drew criticism from civil society and industry figures. John Coleman, legislative counsel for the Foundation for Individual Rights and Expression, said: "No one knows how these companies are picked and why everyone else is excluded," adding that the arrangement "puts too much power in the government's hands with little transparency" [3]. OpenAI CEO Sam Altman said extensive safety testing "is not a bad idea. I just don't like the idea of the government picking the customers"

. OpenAI itself stated that it does "not believe this kind of government access process should become the long-term default," describing its own limited release of GPT-5.6 Sol to government-approved customers as "a temporary step on the path to broader availability"

[3]

[4][.](#src-006)

[6]Cybersecurity and strategy analysts questioned whether the restrictions serve the national security objectives they invoke. Alex Stamos, a Stanford University cybersecurity expert and chief product officer at Corridor, said: "I just want to say that pretty much nobody in the cybersecurity industry believes that there's any factual basis for this action," adding that "if the administration is honest about wanting the US to beat China in the AI race, this is about the dumbest thing they could possibly do" [4]. Kate Koren, an analyst at the Center for Strategic and International Studies and former Commerce Department official, described the order as "a practical interim step, but leaves unresolved the larger issue of how companies can widely release updated models," warning that "the longer there is no system in place for wide release, the more likely China will catch up"

. [3]Outside the United States, coverage framed the restrictions as a shift from hardware-based to model-based export controls with consequences for allied nations. Sina Finance described the policy as an escalation — "从'卡硬件'走向'卡模型'" (from "blocking hardware" to "blocking models") — situating it within the broader US-China technology competition [17]. Lianhe Zaobao characterized the controls as a "fence-style" strategy leveraging technological advantage

. Le Monde reported that the reactivation of Mythos applied to a limited circle of American "cyberdefenders and infrastructure operators" while foreign partners remained excluded

[18]. Euronews's German-language service reported that the EU Commission warned US export controls against Anthropic "dürfen nicht diskriminieren" (must not be discriminatory), framing the matter as one of European technological sovereignty

[13]. Le Figaro reported that European Commission President Ursula von der Leyen called for transatlantic cooperation on AI and insisted that EU citizens and companies should be able to use the best models in security

[14]. Korea JoongAng Daily reported that the controls had prompted allied nations to discuss AI sovereignty and supply-chain strategies [12]. Yeni Şafak described the situation as a "yapay zekada soğuk savaş" (cold war in AI), noting that even foreign engineers at Anthropic were blocked from accessing the models

[21]. [20]Der Spiegel reported the global shutdown of Mythos and Fable 5 following the original US government order as an event without precedent in the AI industry [15]. RTVE described Mythos as having a "doble cara" (double face) — a tool for both defense and potential offense — and noted that European institutions were absent from the approved access list

. The Council on Foreign Relations analyzed Claude Mythos as an inflection point for global security, examining its potential for both defensive and offensive cyber operations and the implications of the consortium access model

[16]. [11]One coverage thread noted a tension within the administration's own posture: Dawn reported that the government intervention was striking for a White House that had otherwise pushed to loosen AI oversight and moved to block states from writing their own AI rules [6].

The current state of affairs leaves Mythos 5 available to the approved domestic organizations while Fable 5 remains restricted and Anthropic's legal challenge to its national security and supply chain designation continues [2] [8]. Both Anthropic and OpenAI have stated that the government-vetted access framework should not become a permanent arrangement

[3][.](#src-004)

[4]