At a press conference at the Conrad Hotel in Seoul's Yeouido district on Wednesday, Anthropic Managing Director of International Chris Ciauri told reporters the company is "very confident that in the coming days, the models will become available again." He was referring to Claude Fable 5 and Mythos 5 — Anthropic's two most capable models, which were pulled offline globally on June 12 after the U.S. government invoked export-control authority over their API access. Ciauri's quote, reported widely on June 18 and June 19, is the first concrete restoration timeline Anthropic has put on record. When asked about "Project Glasswing" — the codename used in reporting that linked the directive to a Korean telecom's suspected China ties — Ciauri declined to comment.
What's actually being negotiated is the technical fix the U.S. demanded. According to David Sacks of PCAST, the administration's pre-export-order offer was binary: patch the specific jailbreak attack — researchers asking the models to "fix this code" on planted-vulnerability samples — or voluntarily pull the models. Anthropic refused both. "Coming days" suggests Anthropic believes it has now closed the technical gap enough to satisfy reviewers, even though security researchers point out that no frontier model can be fully jailbreak-proof, and the deeper issue — using export law against an API — is unresolved.
The timeline matters beyond Anthropic. If Fable 5 and Mythos 5 come back this week, the precedent is that the U.S. government can briefly switch off a frontier model and force a patched release inside two weeks. If they don't, the export-control mechanism remains a live wire that any future administration can pull again, and the EU's Friday announcement of its own sovereign frontier model becomes look less like a luxury and more like a baseline. Either way, the calendar between now and Monday is the most consequential window for AI policy this quarter.
Takeaway for learners: "coming days" is forecasting language, not a commitment. Notice that Ciauri did not say which day, which capability would return first, or which conditions Anthropic had to meet. When a company executive uses imprecise time language during an ongoing regulatory dispute, the imprecision is usually load-bearing. Track what they actually ship against what they hinted — the gap between the two tells you more about the dynamics than the press conference itself.