en.Wedoany.com Reported - After the U.S. government imposed global access restrictions on Anthropic's two AI models, Mythos and Fable, AI startups in Japan and China have begun launching competing products.
Chinese cybersecurity company 360 has launched an AI tool called Tulongfeng (秃鹫风), claiming it can compete with Mythos. The model focuses on cybersecurity and is reportedly extremely capable, making it inaccessible to non-U.S. users.
That same week, Tokyo-based Sakana AI released a model named Fugu (河豚), derived from the Japanese word for pufferfish. The company positions Fugu as a frontier AI model—a high-capability model at the forefront of AI development—with performance comparable to Anthropic's Fable 5 and Mythos Preview. Fugu is also specifically designed for AI agents and can manage access to other models via an API. A Sakana AI spokesperson stated: "Sakana Fugu is a product we have been building since last year." He added that the timing of the release coincided with a particular moment, giving Fugu more attention than expected.
Sakana said Fugu's launch coincided with Anthropic's export ban but was not driven by it. The company promotes Fugu on its website as a frontier-capability model without export control risks. Founded in 2023 by former Google researchers Ren Ito, Llion Jones, and David Ha, Sakana is dedicated to developing generative AI models that are more economical, capable of processing small amounts of data, and optimized for Japanese language and culture. Fugu targets Japanese enterprises and government agencies looking to mitigate risks from tightening export controls. However, the company did not mention that Asia would abandon U.S. AI models. A Sakana spokesperson said: "U.S. models remain important for Asia."
Sakana co-founder and CEO David Ha believes Fugu is more than just a means to capture market share when U.S. competitors are constrained by export rules. He wrote on X: "Orchestrating models is the next frontier, surpassing larger models." Orchestrating models can manage the use of multiple AI models simultaneously, which he sees as the next phase of development.
In China, according to Reuters, 360 launched two AI security tools. Tulongfeng is designed to automatically discover software security vulnerabilities, while Yitianzhen (倚天镇) is used for automated network defense and incident response. 360 founder Zhou Hongyi described this vulnerability-searching AI as a national strategic asset and emphasized the risk of "one-way transparency," where one party has access to advanced vulnerability detection capabilities while the other does not.
Anthropic has achieved tremendous growth. The U.S. AI company reported annual revenue exceeding $47 billion as of May 2026, though it is unclear what proportion comes from enterprise clients in Asia. Since the export restrictions took effect, companies in Tokyo and Beijing have begun offering alternatives to markets that previously relied on Anthropic's models. Local alternatives trained to understand regional languages and nuances are filling this gap.
