Six days into the most aggressive software export action the U.S. government has taken against an American AI lab, Anthropic’s Managing Director of International Chris Ciauri stood at a Seoul press event on June 18 and told the room the company’s two flagship models will be back online soon. “We are very confident that in the coming days, the models will become available again,” Ciauri said, per Korea JoongAng Daily. The event had originally been billed as a Korea expansion launch.
Instead it became damage control for an order whose stated rationale and actual rationale appear to be different documents.
The Commerce Department directive landed at 5:21 p.m. ET on June 12, according to Bloomberg, invoking export-control authority to bar foreign nationals from accessing Fable 5 and Mythos 5. Anthropic responded by pulling both models offline for every user. In a public statement, the company framed the trigger as “a potential narrow, non-universal jailbreak, which essentially consists of asking the model to read a specific codebase and fix any software flaws,” a capability Anthropic said “is widely available from other models (including OpenAI’s GPT-5.5).”
The official story didn’t survive the weekend. The Washington Post reported on June 15 that the White House had separately identified SK Telecom, South Korea’s largest carrier and a $100 million Anthropic investor since 2023, as suspected of ties to China. The administration first asked Anthropic to revoke SK Telecom’s access alone. Around the same time, Amazon researchers reported potential Fable 5 vulnerabilities to the White House, a paper later identified by the Wall Street Journal and shared privately with outside researchers. The combined sequence, per the Post, convinced the administration it couldn’t trust Anthropic to safeguard the technology. SK Telecom has denied any China ties.
Mythos, withheld from general public release because it can find flaws in code previously thought unbreakable, is accessible through Project Glasswing to roughly 150 partners including Google, Nvidia, Microsoft, Apple, Samsung Electronics, and SK Hynix. That partner list is the actual stake. An export order written around a jailbreak that isn’t really a jailbreak is, in practice, a sovereign claim on who gets to use American frontier models.
Katie Moussouris, the Luta Security founder to whom Anthropic privately shared the Amazon paper, told TechCrunch the underlying finding “should never have triggered an export control.” TechCrunch went further, calling the order “a dangerous precedent about how much control [the government] intends to wield over the release of American-made software.”
Ciauri’s confidence in Seoul is the tell. Frontier labs have spent two years arguing their models are too consequential to regulate clumsily. The first time Washington took them at their word, it used the authority sideways, and the lab is now negotiating its product back from a directive that was never really about the product.
Sources
- https://www.anthropic.com/news/fable-mythos-access
- https://www.washingtonpost.com/technology/2026/06/15/how-anthropic-lost-white-houses-trust-then-its-flagship-product/
- https://techcrunch.com/2026/06/15/the-us-governments-anthropic-models-ban-was-never-about-an-ai-jailbreak/
- https://www.bloomberg.com/news/articles/2026-06-13/anthropic-says-us-limits-foreign-access-to-fable-5-mythos-5
- https://www.koreajoongangdaily.com/business/anthropic-confident-of-reenabling-mythos-fable-5-access-in-coming-days-executive/12727522