The global artificial intelligence landscape is undergoing a tectonic shift. In the wake of the Trump Administration’s decisive move two weeks ago to restrict access to Anthropic’s most powerful models—the cybersecurity-focused "Mythos" and its constrained counterpart, "Fable 5"—a void has opened in the Asian market. In a rapid, almost reflexive response, technology firms in Tokyo and Beijing have stepped forward to fill that gap, signaling a move toward "AI sovereignty" that could permanently alter the global tech hierarchy.
As of late June 2026, the narrative of a monolithic, U.S.-led AI ecosystem is fraying. With the unveiling of Tokyo-based Sakana AI’s "Fugu" and the Chinese cybersecurity giant 360’s "Tulongfeng," the message is clear: when the flow of frontier-level technology is interrupted by state-level export controls, local innovation is the inevitable byproduct.
The Chronology of a Disrupted Market
The current volatility can be traced back to the mid-June directive from the U.S. government, which effectively barred Anthropic from providing global access to its flagship cybersecurity models. The decision, aimed at preventing the proliferation of advanced, dual-use cyber-weaponry, sent shockwaves through enterprise sectors in Asia that had begun to rely heavily on Anthropic’s architecture for infrastructure security.
June 15, 2026: The U.S. government formalizes export controls on Anthropic’s Mythos and Fable 5 models, citing national security concerns regarding the potential for these models to identify and exploit software vulnerabilities on a massive scale.
June 22, 2026: Sakana AI, the Tokyo-based startup founded by former Google luminaries, launches "Fugu." While the company insists the release timing was coincidental, it immediately markets the tool as a way to access "frontier capability without the risk of export controls."
June 24, 2026: Chinese cybersecurity firm 360 unveils two strategic AI assets: Tulongfeng, a vulnerability-discovery tool designed to rival Mythos, and Yitianzhen, an automated cyber-defense and incident response system.
June 25, 2026: Diplomatic ripples continue as G7 discussions in Evian highlight the friction between AI-driven national security and the necessity of global technology cooperation.
Strategic Responses: Tokyo vs. Beijing
While both Tokyo and Beijing are reacting to the same U.S. policy, their strategies reflect fundamentally different philosophies.
Sakana AI: The "Collective Intelligence" Hedge
Sakana AI, established in 2023 by industry veterans Ren Ito, Llion Jones, and David Ha, has positioned its new Fugu model not merely as a replacement for U.S. tech, but as an orchestration layer. Recognizing that relying on a single, vulnerable provider—like a U.S.-based AI lab—poses a systemic risk, Sakana is leaning into "agentic" architecture.
"Orchestration Models are the next frontier, beyond bigger models," co-founder David Ha noted on X. His argument is rooted in the idea of a "collective intelligence" hedge. By building a platform that can orchestrate access across multiple, smaller models, Sakana aims to ensure that no single government’s export policy can cripple a Japanese company’s operations overnight.
Sakana’s approach is diplomatic but firm. Through op-eds and G7 appearances, co-founder Ren Ito has urged Washington to reconsider the "hoarding" of AI. Ito argues that AI sovereignty is about having options, not necessarily about total autarky. "We’d characterize the current moment as one of building resilience rather than a permanent realignment toward any one set of players," a company spokesperson told TechCrunch.
360: The "Strategic Asset" Doctrine
In contrast, Beijing-based 360 has adopted a more hawkish tone. For 360 founder Zhou Hongyi, the development of vulnerability-finding AI is not just a business move; it is a national strategic necessity. Zhou has explicitly criticized the concept of "one-way transparency"—the idea that Western powers can maintain access to advanced, cyber-offensive capabilities while others are locked out.
360’s new suite of tools, Tulongfeng and Yitianzhen, represents a direct challenge to the Western monopoly on high-end cybersecurity AI. By integrating these models directly into the domestic security infrastructure, China is signaling that it intends to bypass the U.S. bottleneck entirely, moving toward a self-sustaining ecosystem that is immune to American regulatory shifts.
Supporting Data: The Cost of Dependence
The urgency behind these launches is supported by the staggering growth of the companies they seek to rival. Before the recent restrictions, Anthropic was on a historic trajectory, with its run-rate revenue reported to have crossed $47 billion in May 2026, leading to a valuation nearing $1 trillion ahead of its IPO.
However, this growth was built on a foundation of global enterprise integration. The "export shock" has forced multinational corporations and government agencies in Asia to recalculate their "concentration risk." When a critical piece of software infrastructure—like a vulnerability-scanning AI—can be revoked at the stroke of a pen, the value proposition of "frontier capability" shifts.
The economic implication is clear: even if the U.S. government were to lift its ban tomorrow, the trust deficit has been established. Local models that are optimized for native languages, cultural nuances, and regional regulatory frameworks are gaining a "home-court advantage" that U.S. models may struggle to overcome.
Implications: The End of Global AI Homogeneity?
The emergence of Fugu and Tulongfeng signals that we have entered an era where AI is viewed as a "critical utility" rather than a commercial product. When technology becomes a utility, access becomes a matter of national sovereignty.
1. The Fragmentation of Standards
As different regions develop their own "frontier" models to hedge against export controls, the international standards for AI safety and development will likely diverge. This creates a "splinternet" scenario for AI, where models trained in one jurisdiction may be incompatible or even hostile to the security protocols of another.
2. The Rise of "Agentic" Orchestration
Sakana AI’s focus on orchestration suggests that the next generation of AI will not be defined by the size of the parameter count, but by the ability to manage multiple, disparate models. This is a direct answer to the risk of "platform concentration." Companies will increasingly look for "agnostic" tools that can switch between models based on availability and geopolitical risk.
3. The Diplomacy of Code
The statements made by Ren Ito at the G7 summit highlight that the tech sector is no longer just building software; it is engaging in high-stakes diplomacy. The push for "collective intelligence" over "hoarded capability" is an attempt to define a new norm for AI development—one where collaboration is the primary defense against the weaponization of technology.
Conclusion: A New Equilibrium
The U.S. government’s decision to limit access to Mythos and Fable 5 was intended to secure a strategic advantage. Yet, history suggests that such restrictions often serve as a catalyst for the very innovation they seek to contain. By forcing the hand of developers in Tokyo and Beijing, the U.S. may have inadvertently accelerated the birth of a decentralized AI landscape.
As Sakana AI and 360 continue to iterate on their respective products, the global market is learning a hard lesson: true AI power in the 2020s is not just about the code itself—it is about the resilience of the supply chain. Whether this leads to a "permanent realignment" or a more complex, multi-polar AI ecosystem remains to be seen. What is certain, however, is that the era of relying on a single, U.S.-centric AI provider for global infrastructure is rapidly coming to an end.
The race for the next frontier is no longer just a sprint toward higher intelligence; it is a marathon of strategic autonomy. In this new, fragmented landscape, the winners will be those who can provide not just the smartest models, but the most reliable, sovereign, and flexible intelligence architectures.
