The Ban Created a Market, and Asia Filled It
When the US government blocked Anthropic's Mythos and Fable 5 models from non-American access roughly two weeks ago, the immediate effect was a gap in supply for some of the most capable systems on the market. That gap is now being filled by Asian labs that are not subject to American export rules. Tokyo's Sakana AI launched a model named Fugu, after the Japanese blowfish, and China's 360 shipped a pair of security-focused models. Both arrivals carry an unmistakable subtext aimed squarely at the policy moment.
Sakana describes Fugu as delivering frontier capability without the risk of export controls. A spokesperson called the timing entirely coincidental while admitting the ban brought more attention than the company expected. We are skeptical of the coincidence framing, but the positioning is shrewd regardless. The single most valuable thing a non-American lab can offer right now is durability of access, the promise that a model an enterprise builds on will not be revoked by a foreign government's policy shift.
Sakana's Pitch: Access That Cannot Be Revoked
David Ha, Sakana's chief executive and a co-founder alongside Ren Ito and Llion Jones, put the thesis bluntly. Access to top models can disappear overnight, he said, and collective intelligence is the practical hedge against this concentration of power. Sakana, founded in 2023, was last valued at 2.65 billion dollars in a Series B round in November 2025, and its swarm-of-smaller-models philosophy now reads as a strategic answer to single-vendor dependency as much as a research bet.
Ren Ito struck a more diplomatic note, arguing that US models remain important to Asia and urging Washington to preserve allied access rather than force a permanent realignment. The tension between those two messages is the interesting part. Sakana wants to sell sovereignty and continuity to enterprises spooked by the ban, while also signaling that it does not wish to be cut off from the American ecosystem entirely. That is the needle every ambitious non-US lab is now trying to thread.
China's 360 Reframes Models as National Assets
China's 360, founded by Zhou Hongyi, took a different and more pointed tack. The company released Tulongfeng, aimed at vulnerability discovery, and Yitianzhen, built for cyber defense and incident response, and framed vulnerability-detection AI as a national strategic asset. The choice of domain is telling. Rather than chase general assistants, 360 is staking out the security frontier, where capability has direct geopolitical weight and where a domestic alternative to American tools carries obvious appeal to Chinese buyers and the state.
For Western enterprises, the security framing should be read carefully. AI systems designed to find and exploit software vulnerabilities are dual-use by nature, and a model marketed as a defensive asset in one jurisdiction is an offensive capability in another. The proliferation the export ban was meant to slow is, in this corner of the market, arguably accelerating, with the added complication that the new entrants are being shaped explicitly around national security priorities rather than commercial neutrality.
What This Means for Enterprise Procurement
The strategic lesson for CIOs is that export controls have a counterintuitive second-order effect. By making American frontier models harder to access abroad, Washington has handed Asian labs the most compelling sales argument they have ever had: continuity. Enterprises that operate globally now have a concrete reason to evaluate Sakana, 360, and their peers not as cheaper substitutes but as insurance against geopolitical disruption to their core AI dependencies.
None of this means the new models match Mythos on raw capability, and the burden of proof remains on the challengers to demonstrate frontier performance under independent scrutiny. But the competitive map is changing faster than the policy debate. A control regime designed to preserve an American lead is, in practice, seeding a more multipolar model landscape. Buyers should plan for a world with several credible frontier providers across jurisdictions, and architect their stacks so that no single export decision can take a critical workload offline.
