The global landscape of artificial intelligence is undergoing a seismic shift as major Asian technology firms move to fill the vacuum created by increasingly stringent United States export controls. In a series of rapid-fire developments this week, cybersecurity and AI entities in China and Japan unveiled new frontier-level models specifically designed to compete with—and in some cases, replace—top-tier American AI systems currently restricted by the U.S. government. These launches signal a growing movement toward "AI sovereignty" in the Asia-Pacific region, as nations seek to insulate their domestic industries from the volatility of international trade policy and the risks associated with a reliance on a single geographic source of technology.
The Emergence of Asian Frontier Models
The most significant of these developments came from the Chinese cybersecurity titan 360, which officially introduced its latest AI innovation, Tulongfeng. According to reports from major financial outlets including Reuters, 360 has positioned Tulongfeng as a direct competitor to Anthropic’s Mythos model. Mythos, a cybersecurity-focused AI developed by the San Francisco-based Anthropic, has recently become a flashpoint in the technological cold war between Washington and Beijing. Due to its perceived power in identifying and exploiting software vulnerabilities, the Trump administration recently implemented a ban preventing the model, along with its more restricted counterpart Fable 5, from being accessed by non-American entities.
Simultaneously, the Tokyo-based startup Sakana AI announced the release of Fugu, a frontier AI model named after the Japanese blowfish—a delicacy that is famously potent and requires expert handling. Sakana AI, which has quickly risen to prominence in the Japanese tech ecosystem, claims that Fugu stands "shoulder-to-shoulder" with Anthropic’s Fable 5 and the Mythos Preview. Unlike standard large language models (LLMs), Fugu is specifically architected for "agents," possessing the advanced capability to orchestrate and manage access to other models through their respective APIs. This focus on orchestration suggests a shift in AI development away from singular, monolithic models toward interconnected ecosystems that can survive even if access to one specific provider is severed.
Technical Specifications: Tulongfeng and Fugu
The two new models represent different strategic approaches to the AI market. 360’s Tulongfeng is primarily a defensive and offensive cybersecurity tool. It is designed to automatically discover software vulnerabilities at a speed and scale that exceeds human capability. Alongside Tulongfeng, 360 unveiled Yitianzhen, a secondary tool built to automate cyber defense and incident response. Together, these tools form a comprehensive AI-driven security suite. Zhou Hongyi, the founder of 360, has described these capabilities as "national strategic assets," arguing that in a world of "one-way transparency"—where some nations have advanced vulnerability-detection AI and others do not—those without such tools are left dangerously exposed.
In contrast, Sakana AI’s Fugu is designed for broader enterprise and governmental utility with a focus on cultural and linguistic nuance. Sakana AI specializes in creating generative AI models that are optimized for the Japanese language and culture, often utilizing smaller, more efficient datasets than their American counterparts. Fugu’s primary innovation lies in its orchestration capabilities. By acting as a central hub that can coordinate tasks across various other AI models, Fugu provides a layer of redundancy. If a U.S.-based model becomes unavailable due to new export regulations, Fugu can theoretically reroute its processes to alternative models, ensuring business continuity for Japanese firms.
The Catalyst: US Export Restrictions on Anthropic
The timing of these releases is inextricably linked to the U.S. government’s recent regulatory actions. Approximately two weeks ago, the U.S. Department of Commerce, under the direction of the Trump administration, issued an order preventing Anthropic from granting global access to its Mythos and Fable 5 models. The rationale cited by Washington centered on national security concerns, specifically the fear that these high-level models could be used by adversarial states to develop sophisticated cyber-weapons or disrupt critical infrastructure.
This ban has had an immediate cooling effect on the global AI market. For many Asian enterprises that had integrated Anthropic’s technology into their workflows, the sudden loss of access served as a wake-up call regarding the risks of technological dependency. Anthropic, which had been on an unprecedented growth trajectory with a run-rate revenue crossing $47 billion in May 2026 and a valuation nearing $1 trillion, now faces a significant hurdle in maintaining its dominance in the Asian market. While it is unclear what percentage of Anthropic’s revenue is derived from Asian enterprise customers, the emergence of Tulongfeng and Fugu suggests that the vacuum is being filled rapidly by local players.
Sakana AI: Japan’s Strategic Hedge
While Sakana AI’s spokesperson maintained that the timing of Fugu’s release was "entirely coincidental," the company has not shied away from the marketing opportunities presented by the U.S. ban. The company’s website now prominently features advertisements for "frontier capability without the risk of export controls." This positioning is a direct appeal to Japanese businesses and government agencies that are increasingly wary of being caught in the crossfire of U.S.-China trade tensions.
Founded in 2023 by former Google researchers David Ha and Llion Jones, along with former Mercari executive Ren Ito, Sakana AI has positioned itself as the standard-bearer for Japanese AI. Jones was one of the co-authors of the seminal "Attention Is All You Need" paper, which introduced the Transformer architecture that underpins almost all modern AI, including ChatGPT. This pedigree has allowed Sakana to raise $135 million in a Series B funding round at a $2.65 billion valuation, specifically to build models tailored for the Japanese market.
David Ha, the CEO of Sakana AI, emphasized that Fugu is more than just a opportunistic "land grab." In a statement on the social media platform X, Ha argued that "orchestration models are the next frontier, beyond bigger models." He noted that relying on a single provider for national infrastructure is a systemic risk that the recent export controls have made impossible for world leaders to ignore. "Access to top models can disappear overnight," Ha wrote. "Collective intelligence is the practical hedge against this concentration of power."
360 and China’s National Strategic Asset
In China, the response to U.S. restrictions has been even more assertive. 360’s Zhou Hongyi has been vocal about the necessity of domestic alternatives to Western AI. By framing Tulongfeng as a national strategic asset, 360 is aligning its corporate goals with Beijing’s broader objective of technological self-reliance. The concept of "one-way transparency" mentioned by Zhou highlights a fundamental fear in the Chinese tech sector: that the U.S. will use its lead in AI to create a permanent informational and security advantage.
The launch of Tulongfeng and Yitianzhen is seen by analysts as a declaration that China will not remain passive as the U.S. restricts access to frontier models. Unlike Sakana, which positions its products as a "hedge" to complement existing systems, 360 is positioning its AI as a complete replacement for restricted American tools, specifically in the high-stakes arena of cybersecurity.
Chronology of the AI Export Crisis
To understand the speed of this realignment, one must look at the condensed timeline of events over the past few months:
- May 2026: Anthropic announces its run-rate revenue has hit $47 billion, fueled by global enterprise adoption of its Claude and Mythos series.
- Early June 2026: The Trump administration expresses formal concerns regarding the "dual-use" capabilities of Anthropic’s Mythos model in cyber warfare.
- June 10, 2026: The U.S. Department of Commerce issues a formal ban on the export of Anthropic’s Mythos and Fable 5 models to non-U.S. entities.
- June 17, 2026: At the G7 summit in Evian, France, AI access and export controls become a central topic of diplomatic tension. Sakana AI co-founder Ren Ito attends, advocating for preserved access for allies.
- June 22, 2026: Sakana AI launches Fugu in Tokyo, marketing it as an export-control-proof alternative.
- June 24, 2026: 360 unveils Tulongfeng in China, claiming parity with Anthropic’s restricted cybersecurity models.
The Economic Fallout: Anthropic’s Trillion-Dollar Ambition at Risk
The financial implications of these developments for U.S. AI labs are profound. Anthropic’s trajectory toward a trillion-dollar valuation was predicated on its ability to capture the global enterprise market. By effectively ceding the Asian market—at least in the high-performance and cybersecurity segments—the U.S. government may be inadvertently hobbling its own domestic champions.
Industry analysts suggest that once an enterprise switches its infrastructure to a local alternative like Fugu or Tulongfeng, the "switching costs" to return to a U.S. model are prohibitively high. This is particularly true for models trained on local datasets and optimized for specific linguistic nuances. Even if the U.S. ban were to be lifted in the future, the trust deficit may prove permanent.
The Rise of Orchestration and Agentic AI
A key technical takeaway from the launch of Fugu is the rising importance of "agentic" AI. While the industry has been obsessed with parameter count and raw processing power, Sakana AI’s focus on orchestration suggests that the future of AI utility lies in the ability of models to act as managers. Fugu’s ability to call upon other APIs means it can use a variety of specialized models—some local, some open-source, some international—to complete a complex task. This modular approach is inherently more resilient to geopolitical disruptions than a model that attempts to do everything within a single, closed-source black box.
Geopolitical Implications: Toward AI Sovereignty
The actions of Sakana AI and 360 reflect a broader trend toward "AI sovereignty." Nations are beginning to realize that AI is not just a commercial product, but a foundational utility similar to electricity or telecommunications. Ren Ito, co-founder of Sakana AI, elaborated on this in an op-ed for Project Syndicate, urging the U.S. to prioritize access for its allies. He argued that AI should be developed collectively rather than hoarded, warning that current policies might drive a permanent wedge between the U.S. and its technological partners.
For Japan, the goal is to remain a key player in the global AI ecosystem while ensuring its domestic industry is not paralyzed by decisions made in Washington. For China, the goal is more adversarial: to achieve parity and eventually superiority in a technology that will define 21st-century power.
Conclusion: The New Global AI Equilibrium
The emergence of Tulongfeng and Fugu marks the end of the era of uncontested U.S. dominance in frontier AI. While American models like those from OpenAI and Anthropic remain highly influential and technologically advanced, the introduction of viable Asian alternatives creates a multipolar AI world.
As export controls continue to drag on, the "Splinternet"—the division of the internet and its underlying technologies into regional blocs—appears to be extending into the realm of artificial intelligence. Whether this results in a more resilient global tech ecosystem through "collective intelligence" or a more fractured and dangerous one remains to be seen. However, one fact is clear: the gap left by U.S. restrictions is being filled with remarkable speed, and the new players in Tokyo and Beijing are already establishing deep roots in the market.
