Asian AI labs ship Mythos-style cyber models as Anthropic export ban persists

Category: AI
Status: developing
Sources: TechCrunch, HN discussion
Summary: TechCrunch reported 2026-06-27 that at least two non-US labs launched security-focused models into the gap left by the US foreign-access ban on Anthropic Mythos 5 and Fable 5: Tokyo-based Sakana AI released Fugu, an agent-oriented model it claims stands alongside Fable 5 and Mythos Preview and can orchestrate other models through their APIs, and Chinese security firm 360 unveiled Tulongfeng for vulnerability discovery. The capability claims are the vendors' own and are not independently verified.
Why it matters: Export controls on a single US cyber-capable model are prompting local substitutes, weakening the controls' practical effect and spreading vulnerability-discovery capability across jurisdictions.
Follow-up: Watch for independent benchmarks of Fugu and Tulongfeng, weight or API availability, and whether the Mythos 5 US clearance (2026-06-26) blunts demand for alternatives.