Sakana AI and 360 launch Mythos-like models amid U.S. export controls on Anthropic’s Mythos

Tokyo-based Sakana AI, co-founded in 2023 by former Google researchers Ren Ito, Llion Jones, and David Ha, launched Fugu, a frontier AI model described as comparable to Anthropic’s Mythos and Fable 5. The company positions Fugu as a model designed for agent orchestration, with the ability to coordinate access to other models via their APIs. Sakana’s spokesperson stated the model’s release was not timed to coincide with the U.S. export ban but noted the timing brought additional attention to the product.

Sakana’s leadership emphasized the model’s technical merits and its alignment with the company’s broader strategy of delivering affordable generative AI optimized for Japanese language and cultural contexts. David Ha, co-founder and CEO, framed agent orchestration as the next frontier in AI development, arguing that reliance on a single provider for national infrastructure poses risks, particularly in light of recent export controls.

China’s cybersecurity firm 360 reportedly unveiled Tulongfeng, an AI tool designed to automatically discover software vulnerabilities, and Yitianzhen, a system built to automate cyber defense and incident response. 360’s founder, Zhou Hongyi, described vulnerability-finding AI as a national strategic asset and warned of the risks of 'one-way transparency,' where some actors gain access to advanced capabilities while others do not.

The product launches by Sakana and 360 follow a U.S. export order issued two weeks prior to the article’s publication, which banned Anthropic from providing global access to Mythos and its restricted version, Fable. The order has created an opening for regional AI providers to position alternatives that avoid U.S. export restrictions.

Sakana’s leadership acknowledged the continued importance of U.S. AI models in Asia but framed the current moment as an opportunity to reduce exposure to tightening export controls. Ren Ito, co-founder, reiterated this view in an op-ed published in Project Syndicate, urging the U.S. government to prioritize preserving access for allies and advocating for collaborative AI development rather than hoarding advanced technologies.