Washington’s latest move in the AI race looks less like fine‑tuned regulation and more like yanking the power cord: Anthropic’s flagship models Fable 5 and Mythos 5 have gone dark worldwide after a U.S. order aimed only at foreigners.
Security hawks vs. open‑AI advocates
From the government side, this is framed as classic export control in a high‑stakes tech rivalry. U.S. authorities ordered Anthropic to “block foreign users from accessing” its “most advanced AI models,” prompting the company to comply and cut international customers off from its top systems.1 Another account stresses that Fable 5 and Mythos 5 were disabled following an export‑control order that arrived just days after the models’ launch, amid fears that users may have “bypass[ed] the models’ safety controls” to probe software vulnerabilities.2
To officials, this is a pre‑emptive national‑security firewall: treat frontier AI like sensitive dual‑use tech, cordon it off from potentially hostile states, and accept collateral damage to global access.
Civil libertarians and critics cry overreach
Opposition outlets cast the same facts as a sweeping and clumsy ban. The U.S. Department of Commerce is described as having barred Anthropic from giving foreigners access to “the company’s most advanced AI models, Mythos 5 and Fable 5,” leading to their shutdown “for all users.”3 Another report notes that authorities demanded access be cut not only to foreign customers but also to foreign Anthropic employees “both within and outside the US,” forcing the firm to disable the models entirely to stay compliant.4
Here, the story isn’t prudent controls but an overbroad national‑security dragnet triggered by a single alleged jailbreak technique that Anthropic itself reportedly considers minor and already replicable by other public models.24
Same shutdown, different narrative
Both sides agree on the outcome: frontier systems abruptly pulled offline worldwide. Where they diverge is the spin. For Washington, it’s a necessary shield in a dangerous new era. For critics, it’s a warning shot that AI governance is drifting toward secrecy, fear—and digital protectionism.