Asian AI startups fill gap left by Anthropic's US export ban

The Trump administration's ban on Anthropic's advanced Mythos and Fable 5 models outside the US is triggering an immediate regional response. Japanese startup Sakana AI this week launched Fugu, billed as a frontier model that matches Anthropic's restricted offerings, while China's 360 unveiled Tulongfeng—a vulnerability-detection tool designed to compete with Mythos's cybersecurity focus.

Sakana's release, timed just as the export controls took effect, is being framed as a hedge against concentration risk rather than a permanent US replacement. Co-founder David Ha characterized Fugu as an "orchestration model" designed to coordinate access across multiple AI systems, reducing reliance on any single provider. The company's spokesperson stressed that US models remain strategically important to Asia, but the timing—combined with Sakana's marketing promise of "frontier capability without export controls"—highlights how quickly geopolitical friction translates into business opportunity.

China's 360, by contrast, is playing a longer game. Founder Zhou Hongyi cast vulnerability-detection AI as a national asset, signaling that access to advanced security tools represents a strategic vulnerability. The move reflects broader concern about asymmetric information asymmetries in AI capability.

Both launches underscore a shift already visible in Asia's AI infrastructure: local models trained on regional languages and use cases are moving from nice-to-have to essential, especially as US policy uncertainty makes foreign reliance risky. Anthropic, which reported run-rate revenue exceeding $47 billion in May, now faces a window in which it's unclear whether Asian customers will return even if the ban lifts.