Asian AI firms launch Mythos-style models amid Anthropic export ban

A Chinese cybersecurity firm named 360 recently launched Tulongfeng, an AI tool it says can rival Anthropic’s Mythos. The U.S. government’s export restrictions on Mythos and its restricted variant, Fable 5, have limited access to these models for non-Americans. At the same time, a Tokyo-based company called Sakana AI introduced Fugu, a model it claims can compete with Anthropic’s offerings. This tool is designed for agents and connects to other models through APIs. The U.S. ban on Mythos and Fable 5 was enacted under the Trump administration, reflecting broader concerns about the potential misuse of advanced AI capabilities by non-U.S. entities.

The U.S. export restrictions, which took effect two weeks ago, have created a gap in the AI market. Sakana AI’s representative told TechCrunch that the release of Fugu was not planned in response to the restrictions. However, the company has used the situation to promote its model as offering “frontier capability without the risk of export controls.” The research behind Fugu was presented earlier this year at ICLR, a conference focused on machine learning. The spokesperson elaborated that the company had been developing Fugu since last year, emphasizing that the timing of its launch coincided with the export restrictions but was not a direct reaction to them. The model’s ability to “orchestrate access to other models through their APIs” was highlighted as a key differentiator, allowing users to integrate Fugu with existing systems without relying on a single provider.

Sakana AI was co-founded in 2023 by former Google researchers Ren Ito, Llion Jones, and David Ha. The models developed by the company are tailored for small datasets and adapted for Japanese language and culture. While Fugu is aimed at Japanese businesses and government agencies, it does not indicate a lasting move away from U.S. AI models in the region. Sakana’s focus on optimizing its models for small datasets reflects a broader strategy to address the unique challenges of deploying AI in environments where data availability is limited. This approach has been particularly relevant in Japan, where the need for localized, culturally subtle AI tools is pronounced.

Ren Ito, one of the co-founders, wrote in an article that the U.S. should focus on keeping AI accessible to allies. He said AI should be a shared technology rather than a tool for exclusive control. David Ha, who leads Sakana AI, described Fugu as more than a temporary solution to the export restrictions. He emphasized the importance of “orchestration models” to manage multiple AI systems, reducing dependence on a single provider. His comments on X showed the strategic value of Fugu, framing it as a step toward a future where AI systems can collaborate rather than compete, ensuring resilience against disruptions caused by geopolitical tensions or regulatory shifts.

360’s Tulongfeng and Yitianzhen are built for cybersecurity tasks like finding vulnerabilities and responding to incidents. Zhou Hongyi, the founder of 360, called AI that detects vulnerabilities a key national resource. He warned about “one-way transparency,” where some groups gain access to advanced tools while others do not. This concept, which Zhou emphasized during a Reuters interview, highlights the risks of uneven AI adoption, particularly in a domain like cybersecurity where asymmetries in capability can have significant strategic implications. The tools developed by 360 are not merely technical innovations but are framed within a broader narrative of national security and technological sovereignty.

Anthropic’s revenue reached a run-rate of $47 billion in May 2026, although it’s unclear how much of this comes from Asian companies. The export restrictions have allowed local alternatives to emerge, some better suited to regional languages and needs. These models might continue to serve important roles even if U.S. companies regain access in the future. data labeling efforts in the region could benefit from such tools as they expand their capabilities. The emergence of Fugu and Tulongfeng shows a growing trend of regional AI innovation, driven by the need to address specific local challenges and avoid overreliance on foreign technologies. Even as U.S. firms like Anthropic continue to dominate the global AI setting, the development of localized alternatives suggests a long-term shift in the balance of power within the industry.