Asian AI startups are introducing models similar to Mythos as the export ban from Anthropic continues.

      **TL;DR** Sakana AI and 360 Security from China have introduced AI tools as alternatives to Anthropic's suspended Mythos and Fable models. Sakana AI, based in Tokyo, launched Fugu, an orchestration model it claims performs on par with Fable 5. Meanwhile, Beijing's 360 Security debuted Tulongfeng, a tool for discovering vulnerabilities that it asserts can compete with Mythos. These developments occur as the US government's export ban on Anthropic’s models continues without resolution.

      Sakana's strategy is distinctive; instead of developing a new model from the ground up, they created a seven-billion-parameter orchestrator designed to allocate tasks among various external models. Fugu manages and coordinates these models collaboratively, with the company asserting it can match the performance of much more expensive training systems.

      Founded in 2023 by Llion Jones, a co-author of the Transformer paper from Google, David Ha, a former researcher at Google Brain, and Ren Ito, a previous Japanese diplomat, Sakana secured $135 million in a Series B investment in November 2025, positioning its valuation at nearly $3 billion. Its research on multi-model orchestration was presented at ICLR this spring.

      A Sakana representative informed TechCrunch that the timing of the launch was merely coincidental, though the company has taken advantage of the situation, promoting “delivering frontier capability without the risk of export controls” on its website. Co-founder David Ha emphasized the product's necessity, pointing out that relying on a single provider for national infrastructure is now a risk made apparent by the export ban.

      360 Security’s approach was more straightforward. At the ISC AI 2026 cybersecurity conference in Beijing on Tuesday, founder Zhou Hongyi introduced two tools under the name Yitian Tulong. Tulongfeng focuses on automated vulnerability discovery, while Yitianzhen aims to automate cyber defense and incident response.

      Zhou characterized vulnerability-detecting AI as a vital national asset and cautioned against “one-way transparency,” where some nations can investigate software for vulnerabilities while others cannot.

      Though Zhou acknowledged that Chinese models still trail American ones by about 20 to 30 percent in foundational capability, he insisted that waiting for equal competency was not feasible. The company claims Tulongfeng has identified 3,432 software vulnerabilities, with 105 confirmed by Chinese authorities; however, Reuters noted it could not verify these claims independently.

      These two launches highlight a trend that the Anthropic ban has intensified across Asia. India is considering a $5 billion sovereign AI fund, and AI access was a significant subject at the G7 summit in Evian last week, where Sakana's Ren Ito was invited to participate as a business leader.

      In an op-ed for Project Syndicate, Ito later argued that the US government should focus on maintaining access for its closest allies. He suggested that AI development should be collaborative rather than monopolized and urged Washington to reflect on the implications of treating advanced models as export-restricted weapons instead of communal infrastructure.

      Neither Sakana nor 360 is claiming to entirely replace American advanced AI. Sakana presents Fugu as a safeguard to maintain capability even if access to a single provider is abruptly lost. Ito remarked to TechCrunch that “US models remain important to Asia,” describing the current situation as not yet a complete realignment.

      However, the commercial landscape is evolving faster than diplomatic efforts. Anthropic's run-rate revenue reached $47 billion in May, yet how much of that is generated from Asian enterprise customers remains unknown. What is clear is that two weeks of export restrictions have already led to the emergence of two tangible competitors, positioned in both an allied capital and a rival’s, both highlighting a concern illuminated by Washington’s ban: reliance on American AI poses risks that no level of capability can mitigate.