Asian AI Startups Rise as Anthropic’s Mythos Faces Export Bans

As U.S. export restrictions limit global access to Anthropic’s most powerful models, Asian tech players are rapidly filling the vacuum. From Tokyo to Beijing, new frontier models are emerging to offer local enterprises a way to bypass geopolitical volatility.

The Void Left by Mythos and Fable 5

The landscape of high-end AI changed abruptly two weeks ago when the Trump Administration implemented an export ban on Anthropic’s most advanced models, including the cybersecurity-focused Mythos and its more restricted counterpart, Fable 5. This restriction prevents non-Americans from accessing these specific capabilities, creating a significant gap for international developers and government agencies that rely on frontier-level intelligence.

While Anthropic remains a titan in the industry—with a reported run-rate revenue crossing $47 billion in May 2026—the sudden lack of access has provided a strategic opening for Asian startups to position themselves as reliable, localized alternatives.

Sakana AI’s Fugu: An Orchestration Hedge

In Tokyo, Sakana AI has launched Fugu, a frontier model designed to stand "shoulder-to-shoulder" with Anthropic’s high-end offerings. Founded by former Google researchers including Ren Ito, Llion Jones, and David Ha, Sakana specializes in generative models optimized for Japanese language and culture, making them highly attractive to local businesses and government sectors.

Crucially, Fugu is not just another LLM; it is an "orchestration model." According to CEO David Ha, the next frontier of AI lies in models that can coordinate access to multiple other models via APIs. By building a system that can manage collective intelligence, Sakana aims to provide a practical hedge against the risk of a single provider's access disappearing overnight due to shifting export controls.

China’s 360 Targets Cybersecurity Sovereignty

While Sakana positions its technology as a way to preserve access, China’s cybersecurity firm 360 is taking a more direct approach to technological sovereignty. The firm recently unveiled two specialized AI security tools: Tulongfeng, designed for the automated discovery of software vulnerabilities, and Yitianzhen, built for automated cyber defense and incident response.

360’s founder, Zhou Hongyi, has framed these tools as national strategic assets. He warned of the risks of "one-way transparency," where certain nations possess advanced vulnerability-detection capabilities while others are left blind. This move signals a shift toward building independent, high-stakes AI infrastructure that operates entirely outside the influence of U.S. regulatory shifts.

Why This Matters for the AI Landscape

The emergence of Fugu and Tulongfeng marks a pivotal moment in the "AI arms race." It suggests that the era of U.S. model hegemony may face a permanent challenge, not just through raw compute power, but through cultural nuance and regulatory independence. As local alternatives gain traction, the global AI ecosystem is moving toward a more fragmented, multi-polar structure where "collective intelligence" serves as a defense against geopolitical concentration of power.

Key Takeaways

  • Geopolitical Catalysts: US export bans on Anthropic’s Mythos and Fable 5 have created a high-demand market for non-US frontier models.
  • Orchestration over Scale: Sakana AI’s Fugu focuses on "orchestration," allowing users to manage multiple models to mitigate the risk of relying on a single provider.
  • Strategic Sovereignty: Both Japanese and Chinese firms are developing localized models to ensure national digital infrastructure remains resilient against sudden changes in international trade policy.