Asian startups exploit U.S. AI export controls with Mythos competitors

Export Restrictions Trigger Regional Model Launches

Asian AI startups are moving quickly to fill the void created by U.S. export controls on Anthropic’s flagship models. According to TechCrunch AI, Chinese cybersecurity firm 360 unveiled Tulongfeng on June 26, positioned as a competitor to Anthropic’s Mythos, while Tokyo-based Sakana AI launched Fugu the same week, claiming capability parity with Mythos and Fable 5. The timing coincides with a U.S. government ban preventing Anthropic from selling these models to non-American users—a restriction that has created both market pressure and opportunity for regional vendors.

Sakana’s Hedging Strategy: Calling the Ban Temporary While Capitalizing on It

Sakana’s messaging presents a deliberate tension. According to TechCrunch AI, a Sakana spokesperson told the publication that Fugu’s release was “entirely coincidental,” yet the company’s website advertises the model as “delivering frontier capability without the risk of export controls.” Sakana co-founder Ren Ito and CEO David Ha, both former Google researchers, co-founded the company in 2023 with a focus on models optimized for Japanese language, small-dataset performance, and cost efficiency. Fugu, according to TechCrunch AI, is designed for agent orchestration—coordinating access to multiple models via APIs—positioning it as a multi-model platform rather than a direct Mythos clone.

The spokesman’s statement that Fugu’s research was presented at ICLR in spring 2026 and has been in development since 2025 undercuts any narrative of rushed competitive response. Yet Sakana is explicitly targeting Japanese enterprises and government agencies “looking to reduce their exposure to tightening export controls,” per TechCrunch AI, directly exploiting the regulatory uncertainty created by the ban.

The Broader Regional Resilience Question

Sakana’s public statements resist framing this as a permanent decoupling from U.S. AI. According to TechCrunch AI, the company spokesperson said “U.S. models remain important to Asia,” echoing remarks Ito made at the G7 summit in Evian the week prior, where export controls dominated discussions. Ito argued in a Project Syndicate op-ed cited by TechCrunch AI that the U.S. should “preserve access” for allies and cautioned against hoarding AI technology.

This rhetorical restraint suggests Sakana and other regional vendors are positioning themselves as tactical alternatives during a period of regulatory flux, not as permanent replacements. However, the existence of credible competitors changes procurement calculus for Asian enterprises: they now have domestically controlled options if export restrictions persist beyond the current moment.

Why This Matters

For Japanese enterprises and government agencies currently licensed to Anthropic’s models, the emergence of Fugu creates a switching-cost calculation. Migrating to a regionally hosted alternative eliminates policy exposure but requires retraining teams on a new interface and potentially accepting reduced model capability in non-Japanese tasks. Sakana’s orchestration layer—allowing coordination across multiple models—is designed to lower switching costs by letting teams use both U.S. and regional models simultaneously. If the export ban extends beyond six months, the accumulated licensing costs of exceptions (if available) versus the one-time migration cost to Fugu tips the calculation toward regional options. Regional vendors are, in effect, building supply resilience that didn’t exist before the restrictions, reducing the leverage of future U.S. export controls.