US Government Orders Anthropic to Restrict Access to Advanced AI Models

US Government Orders Anthropic to Restrict Access to Advanced AI Models The sudden shutdown of Anthropic’s most advanced AI models has turned a technical dispute over safety into a geopolitical test of how far governments will go to control frontier AI.

In April, Anthropic unveiled Mythos, an AI system so capable at writing and analyzing code that the company warned it could pose a “global cybersecurity threat” if widely deployed. Access was limited to a small circle of vetted defenders, while a safer, public-facing derivative called Fable 5 followed in early June.

On June 12, the Trump administration issued an unprecedented export-control order directing Anthropic to block non‑Americans from both Fable 5 and Mythos 5, the latest iteration of the model family. Citing unspecified national security concerns, the White House instructed the company to ensure the systems could not be used by foreign nationals, even inside the United States. Because Anthropic could not technically distinguish foreign from domestic users—many of its own staff are foreigners—it pulled both models offline for everyone within about 90 minutes, according to some accounts.

Reports suggest the move followed two alarms: US officials’ anxiety over Mythos access granted to a South Korean telecom they suspected of ties to China, and Amazon researchers’ claim to have found a way around Fable 5’s safeguards, a concern CEO Andy Jassy reportedly escalated to the White House. Anthropic has disputed that this amounted to a broad “jailbreak,” calling it a narrow, patched issue.

Critics say the ad hoc export order exposes “outdated regulation” and an “uncertain and unstable stage in AI governance,” warning that opaque crackdowns are unsustainable for such a fast‑moving technology. Cybersecurity experts argue the directive is counterproductive because it strips defenders of advanced tools; an open letter urged Trump to revoke the order, calling it “dangerous” to remove Mythos‑level capabilities from US networks.

Supporters of strict controls counter that Washington is right to treat frontier AI like a strategic asset. The episode is being framed as a first real test of whether export controls can contain powerful models, much as governments once tried with encryption and spyware—efforts history shows “don’t work” cleanly but do shape global power dynamics. Some commentators argue AI has “granted America vast new power,” casting the order as proof that the US government is now gatekeeper to both leading models and most of the computing resources behind them.

Abroad, the intervention is read as a “warning shot” to other democracies. European politicians have called the ban a “wake‑up call” to build independent AI capacity, even as they weigh competition from cheap, capable Chinese open‑source systems. Analysts worry that if allies fear Washington can abruptly cut access, they may pivot to non‑US providers, undercutting American leadership in AI.

For Anthropic, the immediate fallout is a scramble to restore access under rules “nobody understands,” and a political fight over whether its own aggressive rhetoric about AI risks helped “talk itself into an AI export ban.” Yet investors and industry watchers note that being treated as dangerous, indispensable infrastructure can also burnish the brand—even as it leaves developers locked out and the broader world unsure who will control the next wave of AI.

Continue reading https://foxvector.com/stories/019ef6ce-f7bc-23db-73e0-138724d745be