US Government Restricts, Then Partially Eases Access to Anthropic's New AI Models

US Government Restricts, Then Partially Eases Access to Anthropic’s New AI Models The fight over Anthropic’s most powerful AI models has turned into a live test of how far governments will go to treat AI as sensitive national-security hardware.

In early June, Anthropic introduced Claude Mythos 5 and Claude Fable 5, describing Mythos as “our most capable model for cybersecurity and biology research” and restricting it to a “small group of vetted partners” because it could be used “both for good and for harm.” Fable 5, built on the same underlying model, was pitched as a Mythos‑level system for “the hardest knowledge work and coding problems,” with safeguards that route sensitive cyber and biology queries to a weaker model instead.

Soon after launch, U.S. officials concluded Mythos could uncover vulnerabilities even in “highly sensitive and supposedly secure U.S. government computer systems,” helping tip AI policy fully into a “national security issue” and an “AI-driven state of exception,” in which measures are “reactive, unpredictable, and often seen as draconian.” On June 12, Washington imposed an export-control order that forced Anthropic to disable access to both Mythos 5 and Fable 5, even for many existing customers and foreign-national staff, stunning the industry. A Business Insider headline captured the shock from users suddenly cut off: “Anthropic’s Mythos 5 Gets a Limited Carveout from US Restrictions,” noting that access had been halted under an order that “barred access … by foreign nationals.”

Backlash was swift. Legal-tech startup Legion sued the U.S. government over “taking away Anthropic’s new model,” challenging the order that required keeping Fable 5 and Mythos 5 away from foreign nationals. Commentators warned that Washington’s ad hoc actions were turning into “Anthropic’s Mythos mess,” as negotiations with the Trump administration dragged on with no clear rulebook for future frontier launches.

Allies and rivals abroad watched closely. One Financial Times podcast noted that when “the US government banned a top AI lab from exporting its newest models … the world took notice,” as export controls on Mythos and Fable “locked most of the world out” and pushed countries to seek “sovereign control over AI.” In Asia, startups rushed to fill the gap. TechCrunch reported “Asian AI startups launch Mythos-like models as Anthropic’s export ban drags on,” highlighting Tokyo-based Sakana AI’s Fugu and China’s 360 Security with Tulongfeng, both marketed as alternatives “without the risk of export controls.” A separate report described how these tools were positioned “as alternatives to Anthropic’s suspended Mythos and Fable 5 models,” underlining a broader shift toward independent regional capabilities.

Chinese AI labs moved aggressively as well. The Verge reported that China’s Zhipu AI, via its Z.ai brand, released open‑weight GLM‑5.2, with some researchers claiming it “matches Mythos in certain bug-finding and cybersecurity scenarios,” a level of advancement “particularly concerning to the US government,” because the model “can be downloaded and run by anyone.” Critics on X argued that clamping down on open models at home was self-defeating. One retweeted post complained: “You can ban them and make it impossible for US companies to use them, but this won’t stop A) global open model progress B) bad actors using them. So what exactly is gained by banning open models, including those from China?”

As pressure mounted, U.S. regulators and Anthropic sought a compromise. On June 26, Commerce Secretary Howard Lutnick informed the company that license terms were being revised after Anthropic had “worked with the U.S. government to address risks associated with Mythos 5 and Fable 5,” efforts he said had “yielded significant progress.” Axios reported that “Anthropic’s Mythos is coming back for a select group of entities approved by the U.S. government,” noting that a license would “no longer be required” to export Mythos 5 to organizations listed in an annex and their foreign-national employees, though “the letter does not change restrictions on Fable 5.”

Anthropic publicly welcomed the partial reprieve, saying the government had greenlit limited access to Mythos 5 for U.S. organizations that “operate and defend critical infrastructure,” while stressing that “we’re continuing to work with the government to expand access to Mythos 5 and make Fable 5 available for general use again.” The Verge framed it as “Anthropic’s Mythos 5 is back,” but “somewhat, for a select group of organizations,” with export controls formally still in place for everyone else.

Within hours, more details emerged. TechCrunch reported that the Trump administration now allowed Anthropic’s Mythos 5 “to be used by more than 100 US companies, agencies,” including their non‑American employees and Anthropic’s own foreign staff, after Lutnick wrote that “appropriate safeguards are in place to permit certain trusted partners to access the Claude Mythos 5 Model.” The Next Web summarized the shift under the headline “US clears Anthropic to restore Mythos 5 access for trusted cyber defenders,” noting that Fable 5 “remains restricted” and talks are ongoing.

At the same time, Axios reported that Fable 5 itself could be next. A “Scoop” story said the powerful model was “on track to return soon,” with insiders expecting limits could be lifted as soon as the coming week, marking “a thaw in a bitter four-month standoff” between Anthropic and the administration—though the Pentagon and NSA still had to sign off. Some in industry tried to adapt in the interim: one widely shared post joked that “while we eagerly await Fable 5’s return,” teams had kept running alternative frameworks and models to maintain progress.

Across these moves, analysts argue the deeper pattern is clear. A detailed newsletter on “AI Policy as a National Security Issue” concluded that “from now on, the most significant factor shaping AI policy … will be AI’s real and projected national security risks,” predicting that U.S. measures will remain “reactive, unpredictable, and often seen as draconian” as Washington positions itself as a global gatekeeper for frontier AI. The Verge captured the unease from the tech side under the blunt headline: “Anthropic’s Mythos mess is only getting worse.”

With Mythos 5 partially back online for “trusted” users and Fable 5 still largely dark, the episode has exposed the fault lines between security hawks, AI labs, foreign partners, and open‑model advocates. Whether the U.S. can stabilize this new export‑control regime—without accelerating rival ecosystems in Asia and beyond—remains an open question.

Continue reading https://foxvector.com/stories/019f1320-d213-11d0-73a3-1a628908a221