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 U.S. national security fears and the race for frontier AI came to a head this month when Washington abruptly choked off, then selectively restored, access to Anthropic’s most powerful models, Mythos 5 and Fable 5.

In early June, Anthropic publicly framed Mythos 5 as “our most capable model for cybersecurity and biology research,” restricted to a “small group of vetted partners” because of dual‑use risks. Fable 5, marketed as a “Mythos-level model” for long‑running coding and knowledge‑work tasks, was launched to paying customers on June 9, then advertised as “currently unavailable” after the clampdown. Policy analysts argued that Mythos’ ability to find vulnerabilities in sensitive systems had triggered a new era of “AI policy as a national security issue,” in which real and projected security risks dominate regulatory decisions.

On June 12, the Trump administration issued an export‑control order forbidding foreign nationals from accessing Mythos 5 and Fable 5, even inside Anthropic’s own offices, forcing the company to disable both models globally. The move stunned the industry; one newsletter described an emerging “AI-driven state of exception” in which policy becomes “reactive, unpredictable, and often seen as draconian.” A legal-tech startup, Legion, responded by suing the U.S. government over “taking away Anthropic’s new model,” challenging the foreign‑access ban in court.

As negotiations dragged into a third week, foreign competitors seized the moment. Asian startups launched “Mythos-like models” that promised “frontier capability without the risk of export controls,” explicitly positioning themselves as alternatives to Anthropic’s suspended systems. China’s Zhipu AI released an open‑weight GLM‑5.2 that researchers say “matches Mythos in certain bug-finding and cybersecurity scenarios,” alarming U.S. officials because it can be “downloaded and run by anyone.”

Facing backlash from allies and domestic industry, the administration began to soften its stance. On June 26, Commerce Secretary Howard Lutnick informed Anthropic that “appropriate safeguards are in place” and that Mythos 5 could be redeployed to a “small group of cyber defenders and infrastructure providers,” with a carve‑out for listed U.S. entities and their foreign staff. Business Insider reported a “limited carveout” allowing organizations that “operate and defend critical infrastructure” to regain access, while Fable 5 “remains on ice.”

By June 27, multiple outlets confirmed that Mythos 5 was “back in action — at least, somewhat,” for more than 100 approved companies and agencies, including Anthropic’s non‑U.S. employees, even as export controls stayed in place for everyone else. The Financial Times noted that “Trump administration allows some access to Anthropic’s Mythos,” easing tensions with the lab but leaving concerns over Washington’s “ad hoc” regulatory approach.

Looking ahead, Axios reported that “Powerful Anthropic model, Fable 5, [is] on track to return soon,” with most agencies reportedly satisfied and final sign‑off still pending from the Pentagon and NSA. For now, Anthropic’s flagship workhorse remains dark for customers, even as rival Chinese and Japanese tools, and open‑weight models like GLM‑5.2, demonstrate that strict U.S. controls may be accelerating AI diversification abroad rather than containing it.

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