The landscape of global artificial intelligence is undergoing a rapid transformation as Asian technology firms move to fill the vacuum left by increasingly restrictive U.S. export policies. In a significant escalation of the race for AI supremacy, Chinese cybersecurity giant 360 and the Tokyo-based startup Sakana AI have recently unveiled powerful new models designed to compete directly with high-end American systems. These developments follow a series of aggressive export controls implemented by the U.S. government, specifically targeting Anthropic’s most advanced models, "Mythos" and "Fable 5," which have been barred from use by non-American entities. As the "AI curtain" descends between the West and the East, local players in Beijing and Tokyo are capitalizing on the opportunity to assert technological sovereignty and provide alternatives to U.S.-dependent infrastructure.
The Emergence of Asian Frontier Models
On Wednesday, the Chinese cybersecurity firm 360 officially introduced Tulongfeng, an AI tool specifically engineered for high-stakes cybersecurity applications. According to company reports, Tulongfeng is designed to match the capabilities of Anthropic’s Mythos, a model that has become the centerpiece of recent geopolitical tension. Anthropic’s Mythos is widely regarded as one of the most potent cybersecurity-focused AI models in existence, possessing capabilities for both offensive vulnerability discovery and defensive orchestration. The Trump administration’s decision to ban the export of Mythos and its slightly more restricted counterpart, Fable 5, was predicated on the belief that such tools represent a national security risk if utilized by foreign adversaries.
Simultaneously, in Japan, Sakana AI launched its newest model, Fugu. Named after the Japanese word for blowfish—a delicacy that is famously dangerous if not prepared with extreme precision—Fugu is marketed as a frontier-level model that "stands shoulder-to-shoulder" with Anthropic’s Fable 5 and Mythos Preview. Sakana AI, which has quickly risen to prominence in the Japanese tech ecosystem, emphasizes that Fugu is designed for "agents," allowing it to orchestrate and manage access to various other models through their respective APIs. This move positions Fugu not just as a standalone chatbot or reasoning engine, but as a central nervous system for complex AI operations.
A Chronology of Export Controls and Market Reactions
The current shift in the AI market can be traced back to a series of pivotal events in early to mid-2026. The timeline illustrates a rapid transition from global cooperation to regional protectionism:
- May 2026: Anthropic, the San Francisco-based AI lab, announces a historic growth milestone, with its run-rate revenue crossing $47 billion. The company nears a $1 trillion valuation ahead of a highly anticipated IPO, signaling the massive commercial potential of frontier models.
- Early June 2026: The U.S. government issues an emergency export order. Citing national security concerns, the administration prevents Anthropic from providing global access to its Mythos and Fable 5 models. The ban is absolute for non-American users, including those in allied nations, sparking immediate concern regarding "AI dependency."
- Mid-June 2026: At the G7 summit in Evian, France, AI access and export controls become a central point of contention. Ren Ito, co-founder of Sakana AI, attends the summit and advocates for a more inclusive approach to AI development, warning against the "hoarding" of technology.
- Late June 2026: Within two weeks of the ban, Sakana AI launches Fugu, and 360 unveils Tulongfeng and Yitianzhen. These launches are widely interpreted by market analysts as a direct response to the restricted access to U.S. technology.
While Sakana AI has officially stated that the timing of its launch was "entirely coincidental," the company’s marketing materials have been quick to highlight the benefits of its independence from U.S. regulatory whims. Its website currently advertises the ability to deliver "frontier capability without the risk of export controls," a message that resonates deeply with international businesses wary of relying on American software that could be deactivated "overnight."
360 and the Concept of National Strategic Assets
The move by 360 is particularly notable given the company’s focus on the strategic implications of AI in warfare and espionage. Along with Tulongfeng, which focuses on vulnerability discovery, the firm launched Yitianzhen, a tool built to automate cyber defense and incident response.
Zhou Hongyi, the founder of 360, has been vocal about the necessity of domestic AI in the face of U.S. restrictions. Zhou described vulnerability-finding AI as a "national strategic asset." He warned against a state of "one-way transparency," a scenario where the United States maintains the ability to detect and exploit software vulnerabilities globally while denying other nations the tools to defend themselves or identify similar flaws. By positioning Tulongfeng as a direct rival to Mythos, 360 is signaling that China no longer intends to rely on Western security benchmarks or software suites.
Sakana AI and the Strategy of "Collective Intelligence"
In Tokyo, Sakana AI’s approach is more focused on resilience and "AI sovereignty." Founded in 2023 by former Google researchers David Ha and Llion Jones, along with former Mercari and Stability AI executive Ren Ito, the company has carved out a niche by creating generative AI models optimized for the Japanese language and culture. In late 2025, the company raised $135 million in a Series B round at a $2.65 billion valuation, specifically to build models that could function efficiently on smaller datasets.
David Ha, the CEO of Sakana AI, has argued that the recent U.S. export controls have exposed the inherent risks of a centralized AI infrastructure. He posits that relying on a single provider—especially one subject to the volatile political climate of a foreign superpower—is a strategic liability. Ha has championed the idea of "Collective Intelligence," where orchestration models like Fugu allow users to hedge their risks by coordinating usage across many different models rather than relying on one "monolithic" system.
Ren Ito, writing in a recent op-ed, further emphasized that for allies like Japan, the U.S. government’s first priority should be to preserve access rather than restrict it. He argued that AI should be a technology developed collaboratively among democratic nations to prevent a fragmented and less secure global digital environment.
Economic Implications and Market Realignment
The economic fallout from the export ban on Anthropic’s models is yet to be fully quantified, but the emergence of Tulongfeng and Fugu suggests that the financial impact could be significant. With Anthropic’s revenue run-rate at $47 billion, a substantial portion of its growth was expected to come from Asian enterprise customers in sectors like finance, manufacturing, and cybersecurity.
The introduction of local alternatives threatens to permanently erode the market share of U.S. AI labs in the Asia-Pacific region. Analysts suggest that once a corporation integrates a local model like Fugu or Tulongfeng into its core infrastructure—optimizing it for local languages, cultural nuances, and specific regulatory requirements—the "switching costs" to return to a U.S. model will be prohibitively high, even if export bans are eventually lifted.
Furthermore, the rise of these models indicates a shift toward "sovereign AI," where nations prioritize the development of domestic capabilities to ensure that their critical infrastructure remains functional regardless of international diplomatic tensions. This trend is likely to accelerate as other regions, including the European Union and the Middle East, observe the ease with which the U.S. can sever access to "frontier" technologies.
Technical Analysis: Vulnerability Detection and Orchestration
The technical specifications of these new Asian models reveal a focus on specialized utility over general-purpose conversation. Tulongfeng’s primary function is the automated discovery of "zero-day" vulnerabilities—flaws in software that are unknown to the developers. In the hands of a cybersecurity firm like 360, this tool can be used to patch domestic systems more rapidly. However, the same technology in an offensive context is what originally prompted the U.S. ban on Mythos.
Fugu, on the other hand, represents the "next frontier" of AI: orchestration. Rather than attempting to be the largest model in terms of parameters, Fugu is designed to be the most "intelligent" manager. It can determine which task is best suited for which specific sub-model, effectively acting as a router that optimizes for cost, speed, and accuracy. This architectural choice makes it highly resilient; if one model (such as a U.S.-based API) becomes unavailable, Fugu can theoretically reroute the task to a domestic or open-source alternative without interrupting the user’s workflow.
The Future of Global AI Governance
The simultaneous launch of Tulongfeng and Fugu serves as a clear indicator that the era of U.S. dominance in the AI sector is facing its most significant challenge to date. While U.S. models currently maintain a lead in raw compute and general reasoning, the "localization" of AI in Asia is creating a competitive ecosystem that is better tuned to the specific needs of regional governments and businesses.
As Sakana AI’s spokesperson noted, U.S. models remain important to the Asian market for the time being. However, the current momentum suggests a "realignment" is underway. If the U.S. continues to use AI access as a tool of geopolitical leverage, it may find that the rest of the world simply builds its own tools, leading to a bifurcated tech world where "collective intelligence" replaces the previous reliance on a few Silicon Valley giants. The success of 360 and Sakana AI in the coming months will likely determine whether this shift is a temporary reaction to policy or the beginning of a new, multipolar era in artificial intelligence.
