Asian AI startups launch Mythos-like models as Anthropic’s export ban drags on

Asian AI Startups Launch Mythos-like Models as Anthropic’s Export Ban Drags On

The global AI landscape is shifting rapidly as Asian startups roll out models that closely mirror the capabilities of Anthropic’s Mythos, now inaccessible to non-Americans due to U.S. export restrictions. This development underscores a growing concern among regional tech ecosystems: the risk of geopolitical fragmentation in AI access and the emergence of localized alternatives to U.S.-centric models. As the ban drags on, companies in Asia are accelerating their efforts to fill the vacuum left by Anthropic, offering tools that are not only competitive in performance but also tailored to local needs and regulatory environments.

A Strategic Move in AI Localization

Asian AI startups are leveraging the export restrictions as an opportunity to build models that are not only technically comparable but also culturally and linguistically optimized for their markets. Sakana AI’s Fugu, for instance, is explicitly designed for Japanese language and business contexts, while 360’s Tulongfeng and Yitianzhen focus on cybersecurity applications that are crucial for national infrastructure.

• Fugu is built to work with smaller datasets, a feature that resonates with Japanese enterprises handling localized and niche information.
• Tulongfeng is designed to detect software vulnerabilities automatically, addressing a critical need in cybersecurity.
• Yitianzhen automates incident response, which is vital for maintaining digital resilience in a highly regulated environment.

These models are not just filling a gap—they are redefining the parameters of AI deployment in Asia, where data privacy and regulatory compliance are paramount.

The Push for Collective Intelligence and Orchestration

Sakana AI’s co-founder David Ha has positioned Fugu as a model that can coordinate access across multiple AI systems. This “orchestration” approach is seen as a response to the increasing centralization of AI power, particularly in light of the export controls. The idea is that no single AI model should control critical infrastructure, and that redundancy and collaboration can serve as a safeguard against geopolitical or technological disruption.

• Fugu is designed to work with various models via APIs, allowing for a more distributed AI architecture.
• This strategy contrasts with the reliance on single providers that the export controls have rendered risky for enterprises.
• Orchestration models are gaining traction as a practical solution to the growing concerns around AI dependency.

The push for orchestration models is not just a technical innovation—it’s a strategic response to the geopolitical realities shaping AI today.

A New Era of AI Autonomy in Asia

The emergence of these models signals a broader trend: Asia is no longer content with being a passive consumer of U.S. AI technology. Instead, it is asserting its own path, with startups and national labs developing tools that are not only competitive but also aligned with regional priorities.

• The U.S. government’s export controls have created an opening for local innovation.
• Companies like 360 and Sakana are capitalizing on this moment to build out their AI ecosystems.
• While the U.S. remains a key player in AI development, its dominance is being challenged by a new wave of regional contenders.

For now, the U.S. models are still seen as vital, but the long-term trajectory is clear. As Asian startups continue to refine their offerings and align them with local needs, the global AI landscape is likely to become more multipolar. The future may not be defined by a single dominant player, but by a network of regional powers each asserting their own vision of AI’s role in the world.