The Week the Tap Got Cut
Two weeks ago, the U.S. government issued an export order that effectively locked non-Americans out of Anthropic's two most powerful models: Mythos and its more restricted sibling, Fable 5. No fanfare. No transition period. Just a directive that changed the global AI landscape overnight.
The timing couldn't have been worse for companies that had built their AI strategies around Anthropic's capabilities. Enterprise customers in Tokyo, Seoul, Singapore — they'd been planning their deployments. Now those plans were suddenly obsolete.
But here's what's interesting: the Asian AI ecosystem didn't panic. It launched products.
Within days of the export order taking effect, two startups from completely different strategic backgrounds made moves that would have been unthinkable just a month earlier. One was positioning itself as a hedge against future restrictions. The other was declaring outright competition.
The question hanging over everything: is this a temporary disruption, or the beginning of something permanent?
Sakana's Fugu: Shoulder-to-Shoulder
Tokyo-based Sakana AI unveiled Fugu earlier this week. Named after the Japanese word for blowfish — appropriately dangerous if you don't know what you're doing — the model is Sakana's answer to the export ban.
The company claims Fugu "stands shoulder-to-shoulder with leading models like Anthropic's Fable 5 and Mythos Preview." That's a bold claim for a startup founded in 2023 by former Google researchers Ren Ito, Llion Jones, and David Ha.
But there's something else going on here. Fugu isn't just another frontier model. It's designed for agents — specifically, it can orchestrate access to other models through their APIs. Think of it as a conductor for an orchestra where the musicians might change instruments mid-performance.
David Ha, Sakana's CEO, put it bluntly on X: "Orchestration Models are the next frontier, beyond bigger models." He argued that relying on a single provider for national infrastructure is a risk the export controls made impossible to ignore. "Access to top models can disappear overnight," he wrote. "Collective intelligence is the practical hedge against this concentration of power."
The strategic calculus here is fascinating. Sakana isn't trying to replace Anthropic. It's building insurance.
The company is targeting Japanese businesses and government agencies looking to reduce their exposure to tightening export controls. And while Sakana's spokesperson called the timing "entirely coincidental," the company's website now advertises "delivering frontier capability without the risk of export controls." That's not coincidence. That's capitalizing on a moment.
360's Tulongfeng: No Hedging, Just Winning
If Sakana is playing defense, China's 360 is playing offense.
The cybersecurity firm unveiled two AI security tools: Tulongfeng, designed to automatically discover software vulnerabilities, and Yitianzhen, built to automate cyber defense and incident response. Both are positioned as direct competitors to Anthropic's Mythos.
But 360's founder Zhou Hongyi didn't just launch products. He launched a warning.
According to Reuters, Zhou described vulnerability-finding AI as a "national strategic asset" and flagged what he called the risk of "one-way transparency" — a situation where some actors could access advanced vulnerability-detection capabilities while others cannot.
That's a loaded concept. If you can find vulnerabilities faster than your adversary, you have an asymmetric advantage. And if that capability is restricted by export controls, you've created a permanent gap.
360 didn't respond to requests for comment. But the message is clear: China isn't interested in hedging. It wants parity, or better.
The timing here is deliberate. By launching Tulongfeng and Yitianzhen immediately after the export order, 360 is signaling that China has its own frontier capabilities — and it's not waiting for permission to use them.
The Strategic Calculus: Hedge or Replacement?
Here's where it gets complicated.
Sakana explicitly does NOT characterize this as a permanent shift away from U.S. AI in Asia. "U.S. models remain important to Asia," the company said.
That view is consistent with remarks co-founder Ren Ito made at the G7 summit in Evian last week, where AI access and export controls were central topics. Ito published an op-ed in Project Syndicate urging the U.S. 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."
So what are these Asian labs actually doing? They're building redundancy. If the export ban lifts tomorrow, they'll still have competitive products. If it doesn't — and that's the question — they won't be dependent on U.S. models for their most critical AI infrastructure.
It's the kind of insurance policy you buy when you suspect the house might burn down, even if you hope it won't.
The distinction between Sakana's approach and 360's is telling. Sakana wants to keep the door open for U.S.-Asia AI cooperation. 360 is preparing for a world where that cooperation doesn't exist.
Anthropic at $47 Billion: What Does the Ban Cost?
Anthropic had been on a historic growth trajectory. The lab said its run-rate revenue crossed $47 billion in May 2026.
How much of that depends on Asian enterprise customers? Nobody's saying. But when you're pulling in that kind of revenue, even a small percentage from international markets represents real money.
The export order came two weeks before the TechCrunch article ran. That's a tight window. It suggests this wasn't a gradual tightening — it was a decision, executed quickly.
For a company this size, the ban is less about immediate revenue loss and more about strategic positioning. If Asian markets permanently shift toward local alternatives, Anthropic loses access to some of the fastest-growing AI adoption regions in the world.
The question isn't whether Anthropic can survive without Asia. It's whether it can thrive.
And there's another dimension to this: if the export ban pushes allies toward local alternatives, you might be creating the very capability you're trying to prevent. That's a paradox worth sitting with.
One-Way Transparency: The Real Danger
Zhou Hongyi's concept of "one-way transparency" deserves more attention than it's getting.
In cybersecurity, vulnerability discovery is everything. If you can find bugs before your adversary, you control the battlefield. But if that capability is restricted by export controls, you create a permanent asymmetry.
Imagine a world where only certain countries can access advanced vulnerability-detection AI. The companies and governments in those countries would have a massive advantage. They'd know about your bugs before you do.
That's not just an unfair playing field. It's a national security risk.
360's Tulongfeng is positioned as China's answer to that problem. But it also raises uncomfortable questions: if vulnerability-finding AI becomes a strategic asset, who decides who gets access? And what happens when that access is weaponized?
The export ban didn't just restrict model access. It created a new category of strategic asset — and that's something the policy community needs to think about.
The concept of "one-way transparency" suggests that asymmetric access to vulnerability-detection capabilities could become a permanent feature of the geopolitical landscape. That's a future most policymakers aren't prepared for.
What Comes Next
The next few months will tell us whether this is a temporary disruption or the start of something permanent.
Sakana's spokesperson called Fugu's timing "entirely coincidental," yet the company's website now advertises "delivering frontier capability without the risk of export controls." That's not coincidence. That's capitalizing on a moment.
360 didn't even pretend the timing was accidental. It positioned its tools as direct competitors to Anthropic's most powerful models.
The U.S. government needs to decide what it actually wants here. If the goal is to prevent adversarial access, fine. But if the ban also pushes allies toward local alternatives, you might be creating the very capability you're trying to prevent.
Ren Ito's op-ed in Project Syndicate is worth reading. He argued that AI should be "developed together," not hoarded. Whether the U.S. government listens remains to be seen.
But one thing is clear: the Asian AI ecosystem just proved it can move fast. And that changes everything.
The next few months will be critical. Will the export ban lift? Will Asian labs continue to invest in frontier capabilities? And will U.S. companies find a way to maintain access to these markets?
The answers to those questions will shape the next decade of AI development.