Two Asian companies have recently launched AI offerings as the U.S. government’s export restrictions have limited the global availability of certain Anthropic models. On Wednesday, Chinese cybersecurity firm 360 reportedly unveiled an AI tool called Tulongfeng, which the company says can compete with Anthropic’s Mythos. Earlier the same week Tokyo-based Sakana AI launched a model named Fugu, after the Japanese word for blowfish.
What the companies say and what the models do
Sakana AI says Fugu “stands shoulder-to-shoulder with leading models like Anthropic’s Fable 5 and Mythos Preview.” The startup emphasizes that Fugu is designed for agents and can orchestrate access to other models via their APIs. A Sakana spokesperson told TechCrunch that the model’s release was “entirely coincidental” with the Anthropic export restrictions, while the company’s website promotes delivering “frontier capability without the risk of export controls.”
According to Sakana, work on Fugu began last year and the underlying research was presented at ICLR this spring. Sakana AI was co-founded in 2023 by David Ha and Llion Jones, both with Google backgrounds, together with Ren Ito, who previously held roles at Mercari and Stability AI. The company focuses on affordable generative AI models that perform well with small datasets and are optimized for the Japanese language and culture.
Sakana is targeting Fugu at Japanese businesses and government agencies that want to reduce exposure to tightening export controls, but it does not claim this marks a permanent departure from U.S. AI in the region. A Sakana spokesperson said, “U.S. models remain important to Asia,” echoing co-founder Ren Ito’s comments at the G7 summit in Evian last week. In an op-ed in Project Syndicate, Ren Ito urged the U.S. federal government to prioritize preserving access for close allies and argued that AI should be developed collaboratively rather than hoarded.
David Ha, Sakana’s co-founder and CEO, wrote on X that orchestration models are “the next frontier, beyond bigger models,” describing Fugu as a system intended to coordinate agent usage across multiple models. He added that relying on a single provider for national infrastructure is a risk highlighted by recent export controls: “Access to top models can disappear overnight,” he wrote, calling collective intelligence a practical hedge against concentration of power.
China’s approach: 360’s Tulongfeng and Yitianzhen
While Sakana positions Fugu as a hedge to preserve access rather than a replacement for U.S. models, 360 took a more direct stance. The company reportedly introduced two AI cybersecurity tools: Tulongfeng, designed to automatically discover software vulnerabilities, and Yitianzhen, intended to automate cyber defense and incident response.
The launch carried a strategic message. Reuters reported that 360’s founder Zhou Hongyi described vulnerability-finding AI as a national strategic asset and warned about the risk of “one-way transparency,” where some actors could gain access to advanced vulnerability-detection capabilities while others could not.
Background: the U.S. export order and Anthropic’s revenue
The U.S. order that prevents Anthropic from providing Mythos and the more restricted Fable 5 to non-U.S. users went into effect two weeks before these announcements. Anthropic has said its run-rate revenue crossed $47 billion in May 2026, though it is not publicly known how much of that depends on Asian enterprise customers.
Since the export order took effect, at least two companies — one in Tokyo and one in Beijing — have stepped into the gap. Even if U.S. firms regain trust should the ban lift, local alternatives trained for local languages and cultural nuance are already filling demand.
360 did not respond to requests for comment, according to the reporting.
Why this matters
These developments illustrate how access policies can reshape global AI supply chains quickly: export controls can create near-term market opportunities for local developers, while also raising strategic questions about technological concentration and unequal access to critical capabilities.
