The Vacuum Effect: How Anthropic's Export Ban Opened the Door for Asian AI Startups
Anthropic had been on a historic growth trajectory. The US AI lab said its run-rate revenue crossed $47 billion in May 2026. Then the Trump administration banned Mythos and Fable 5 from reaching non-American users two weeks ago, and that growth story hit a wall it didn't see coming.
At least two companies, one in Tokyo, one in Beijing, stepped into the space Anthropic left behind. And they're not pretending it's charity. They're building products.
Sakana Fugu: the hedge that isn't hiding
Earlier this week Sakana AI, a Tokyo-based startup founded in 2023 by former Google researchers Ren Ito, Llion Jones, and David Ha, launched Fugu. The name is the Japanese word for blowfish. The company says this frontier model "stands shoulder-to-shoulder with leading models like Anthropic's Fable 5 and Mythos Preview."
But here's the thing that makes Fugu different from a typical competitor. It's designed as an orchestration model, meaning it coordinates access to other models through their APIs. David Ha, co-founder and CEO, put it bluntly on X: "Orchestration models are the next frontier, beyond bigger models."
His argument is simple and hard to dismiss. "Access to top models can disappear overnight," he wrote. "Collective intelligence is the practical hedge against this concentration of power."
That's not a pitch about replacing US AI. It's a pitch about not being left with nothing when the access gets cut off. Sakana's website advertises "delivering frontier capability without the risk of export controls," which tells you exactly who they think their customers are.
360 Tulongfeng: China isn't hedging
While Sakana positioned Fugu as insurance, China's 360 wasn't hedging at all. The Chinese cybersecurity firm reportedly unveiled Tulongfeng, an AI tool it says can go head-to-head with Anthropic's Mythos. Alongside it came Yitianzhen, a tool built to automate cyber defense and incident response.
According to Reuters, 360's founder Zhou Hongyi described vulnerability-finding AI as a national strategic asset. He flagged what he called the risk of "one-way transparency," a situation where some actors could access advanced vulnerability-detection capabilities while others could not.
The framing matters. This isn't a startup trying to capture market share. It's a cybersecurity firm framing its AI as a matter of national security, and using the export ban as proof that China needs its own tools.
What the ban changed
A spokesperson at Sakana AI told TechCrunch the release was "entirely coincidental." I'm not sure that's true or false, but it doesn't really matter. The result is the same: local alternatives, trained to better understand local language and nuance, are filling the gap.
Co-founder Ren Ito made his position clear in an op-ed published in Project Syndicate last week. He urged the US federal government to consider that its "first priority should be to preserve access" for America's closest allies. He argued that "AI should not become a technology that is hoarded; it should be one that is developed together."
He also made remarks at the G7 summit in Evian last week, where AI access and export controls were one of the central topics. The message was consistent: "U.S. models remain important to Asia," but the region is preparing for a world where that importance comes with strings attached.
What this means for developers
If you're building AI applications and relying on a single provider, this week should make you uncomfortable. Not because Anthropic is going anywhere specific, but because the principle has been established. Access can be restricted by government order, and when that happens, alternatives appear fast.
Sakana Fugu and 360 Tulongfeng aren't the last models that will launch to fill a geopolitical gap. They're the first. And the next time a ban hits, or an export control tightens, the infrastructure for local alternatives will already be in place.
The practical takeaway: build your systems so swapping models isn't a rewrite. Because the model you depend on might not be the model you're using six months from now.