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 clash between cutting‑edge AI and national security has turned Anthropic’s newest models, Mythos 5 and Fable 5, into test cases for how far governments will go to control frontier systems.

In early June, Anthropic introduced Claude Mythos 5 as its most capable model for cybersecurity and biology research, restricting it to “a small group of vetted partners” because it “could be used both for good and for harm.” Fable 5, a Mythos‑level system tuned for long‑running coding and knowledge‑work tasks, launched days later but with safeguards that route sensitive cyber and bio queries to an earlier model, Opus 4.8.

On June 12, the U.S. government abruptly imposed export‑control rules that forced Anthropic to disable access to both Mythos 5 and Fable 5, including for its own foreign staff, citing unspecified “national security concerns.” Commentators framed the move as part of a broader shift in which “AI policy is treated as a national security issue,” making measures “reactive, unpredictable, and often seen as draconian.” One AI startup, Legion, responded by suing the government over the order that “required Anthropic to keep its models Fable 5 and Mythos 5 away from foreign nationals.”

As the two‑week blackout dragged on, Washington’s allies and smaller AI nations fretted about dependence on U.S. providers, after export controls “locked most of the world out” of using Anthropic’s systems. Asian firms moved to fill the gap: Sakana AI in Tokyo pitched its Fugu model as “delivering frontier capability without the risk of export controls,” while China’s 360 Security and Zhipu AI claimed tools that could rival Mythos in cybersecurity. That, in turn, sharpened U.S. worries about Chinese progress with open‑weight models that “can be downloaded and run by anyone,” potentially empowering malicious actors.

On June 26, after intense negotiations, Commerce Secretary Howard Lutnick notified Anthropic of a “revision to the license requirements.” Mythos 5 could be “redeployed to a small group of cyber defenders and infrastructure providers,” with a license no longer required to export it to entities listed in an annex and their foreign employees. TechCrunch reported that more than 100 U.S. agencies and companies were cleared for access, including non‑American staff and Anthropic’s own foreign employees. Business Insider described it as a “limited carveout” for U.S. critical‑infrastructure operators, while “Fable 5 remains on ice.”

The partial reopening echoed a similar arrangement for OpenAI’s GPT‑5.6 and eased tensions between Anthropic and the Trump administration, even as analysts warned that Washington’s ad hoc export‑control regime is becoming the de‑facto gatekeeper for global AI access. Industry voices questioned whether such bans can meaningfully contain risk in a world where open and foreign models are proliferating: “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,” one researcher argued on X.

Fable 5, marketed by Anthropic as “the next generation of intelligence for the hardest knowledge work and coding problems,” remains largely inaccessible while the Pentagon and NSA continue their review, even as Axios reports insiders expect limits “could be lifted as soon as this coming week.” For now, Mythos 5’s tightly controlled return—and Fable 5’s limbo—expose the unresolved tension between securing powerful AI and ensuring fair, global access to it.

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