US Government Orders Anthropic to Restrict Access to Advanced AI Models The abrupt shutdown of Anthropic’s newest AI models has turned a technical safety debate into a global fight over who controls frontier artificial intelligence.

In early April, Anthropic quietly released Mythos, an extremely capable coding and cybersecurity model, only to limit it to about 150 vetted organizations under an internal program dubbed “Project Glasswing.” Soon after, the company prepared a safer, public derivative: Fable 5, marketed as “the world’s most powerful cybersecurity model,” with added guardrails and filters.

On June 9, Fable 5 went live for general users. Within days, Amazon researchers reportedly found a way to bypass some of its safeguards, a concern that CEO Andy Jassy relayed to the White House. Around the same time, U.S. officials grew alarmed that a South Korean telecom with alleged — and disputed — links to China had access to Mythos via a limited partner program.

On June 12, the Trump administration invoked national security authorities to impose export controls on Mythos and Fable, ordering Anthropic to block foreign nationals both abroad and inside the U.S. Because it could not reliably distinguish foreign users — including some employees — Anthropic instead “pulled the plug” on both models for everyone within roughly 90 minutes of being notified. Commentators noted this was “the first time US export controls have been used to control access to an AI model in this way.”

From Washington’s perspective, the move showcases U.S. leverage over a technology now seen as strategic infrastructure: by one account, the order proved “the American government can decide who may use the world’s most important technology.” Critics, including cybersecurity experts, counter that yanking advanced defense tools from network defenders is “dangerous” and reflects an ad hoc, opaque approach to AI governance.

The backlash has spread overseas. European politicians have called the episode a “wake-up call” for digital sovereignty, warning that sudden U.S. controls could push users toward cheap, capable Chinese open‑source models instead. Analysts also argue that three decades of attempts to choke off encryption and spyware show cyber export controls “don’t work,” questioning why they would succeed with Mythos and Fable.

Anthropic itself says the risks of Fable 5 were already known and “overstated,” and that the response is “disproportionate.” Some suggest the firm helped create its own predicament: Financial Times analysis found Anthropic warned about AI risks far more often than rival OpenAI, prompting accusations that its “fear‑mongering” helped trigger the crackdown.

The dispute has now moved into the courts. Legal tech startup Legion, which relied on Fable 5, has sued the U.S. government over the order that “took away Anthropic’s new model,” arguing that vague export rules have arbitrarily cut off critical tools for foreign customers.

As developers, investors and allied governments watch, the Anthropic case has become a first test of how far America will go in weaponizing access to AI — and whether the rest of the world will accept those terms.

Continue reading https://foxvector.com/stories/019ef962-360b-2345-727e-3c2074a0000b