Asian AI startups launch Mythos-like models as Anthropic's export ban drags on
TLDR Dev ● Covered by 3 sources
US export ban on Anthropic's Mythos and Fable 5 models is boosting rivals. Japan's Sakana and China's 360 just launched competing AI tools.
Two weeks after Washington cut off global access to Anthropic's Mythos and Fable 5 models, the vacuum left behind is getting filled fast. Tokyo-based Sakana AI dropped a model called Fugu this week, and Chinese cybersecurity firm 360 followed with Tulongfeng, a tool it claims can match Mythos on its home turf: finding software vulnerabilities.
Sakana insists the timing was coincidental. The research behind Fugu was presented at ICLR back in the spring, according to a company spokesperson, and the product has been in development since last year. Maybe so, but the marketing doesn't read like an accident. Sakana's own website pitches Fugu as delivering frontier capability without the risk of export controls, which is about as pointed a jab at U.S. policy as a startup can make without naming names.
What makes Fugu interesting isn't just the timing, though. It's built around orchestration, meaning it can call on other models through their APIs rather than trying to be the smartest single system in the room. CEO David Ha framed this as the real frontier now, arguing that leaning on one provider for critical infrastructure is a gamble the export ban just made impossible to ignore. Access to top models can vanish overnight, he wrote on X, and collective intelligence is the practical hedge against that kind of concentration.
Sakana's co-founder Ren Ito made a similar case in an op-ed for Project Syndicate, urging Washington to prioritize preserving access for allies rather than treating AI as something to hoard. That's a hedging strategy, not a declaration of independence from American AI. 360's move in China feels different. Founder Zhou Hongyi described vulnerability-hunting AI as a national strategic asset and warned against what he called one-way transparency, where some countries get the tools and others don't. That's not a company protecting optionality. That's a government-adjacent firm building sovereign capability on principle.
Anthropic, meanwhile, was riding a run-rate revenue of $47 billion as of last May, though nobody outside the company knows how much of that came from Asian enterprise customers now locked out. Whatever the number, the longer this ban drags on, the more local players get a running start building models tuned to local languages and regulatory habits. Even if the export controls lift tomorrow, Anthropic won't just be re-entering these markets. It'll be competing in ones that have already learned to live without it.
My take
Export controls always sound clever in a policy memo and always backfire the same way: they hand your competitors a two-week head start and call it national security. Sakana's hedging language is smart positioning, but 360 framing vulnerability-detection AI as sovereign infrastructure is the real story here — that's the kind of rhetoric that turns a temporary ban into a permanent bifurcation of the AI stack, and once local models are trained on local nuance, American labs don't just walk back in.
Read more about this at: TLDR Dev
