Asian AI startups launch Mythos-like models as Anthropic’s export ban drags on
TL;DR
China’s cybersecurity company 360 unveiled Tulongfeng, an AI tool for automated vulnerability discovery that it says can compete with Anthropic’s Mythos. Tokyo-based Sakana AI launched Fugu, a model aimed at Japanese businesses and government agencies, with an agent-oriented design that can coordinate calls to other models via API. The political opening matters: the U.S. export order still blocks non-Americans from Anthropic’s Mythos and Fable 5. Asian vendors are now pitching lower exposure to export controls.
Nauti's Take
This is not a clean win for Asian startups or a simple loss for Anthropic. It is a trust fracture in the market.
If AI is sold as infrastructure, benchmarks are not enough; access has to be predictable. Sakana’s pitch looks more credible than 360’s because it sounds less like a power play.
Still, every week of export-control uncertainty makes local alternatives feel less like a backup plan and more like the default.
Briefingshow
This shows how export controls can redirect a market as much as protect one. If Asian customers learn that critical AI capabilities can vanish overnight, raw model performance becomes only one factor next to availability, local language fit, compliance, and political resilience.
Sources