The global landscape of artificial intelligence underwent a significant shift in June 2026 as prominent technology firms in China and Japan unveiled advanced AI models designed to fill the vacuum left by tightening United States export controls. Following a recent mandate from the Trump administration that restricted global access to Anthropic’s most powerful cybersecurity-focused models, Beijing-based 360 Security Group and Tokyo-based Sakana AI have launched products they claim rival the capabilities of the banned American technology. This rapid emergence of localized alternatives highlights a growing trend of "AI sovereignty" and suggests that U.S. efforts to maintain a technological edge through restrictive trade policies may be accelerating the development of independent, high-capability AI ecosystems in Asia.
On Wednesday, the Chinese cybersecurity titan 360 Security Group announced the release of Tulongfeng, an artificial intelligence tool specifically engineered for high-level cybersecurity applications. According to company reports and statements from founder Zhou Hongyi, Tulongfeng is designed to match the performance of Anthropic’s Mythos, a model that has become a centerpiece of the current geopolitical tug-of-war over sensitive technology. Mythos and its slightly more restricted counterpart, Fable 5, were recently classified by the U.S. government as too powerful for distribution to non-American entities, citing concerns over their potential use in offensive cyber operations.
The launch of Tulongfeng was accompanied by a second tool, Yitianzhen, which focuses on automated cyber defense and incident response. Together, these tools represent China’s most direct response to date regarding the exclusion of its firms from the upper tiers of the American AI market. Zhou Hongyi described the ability to find and patch software vulnerabilities via AI as a "national strategic asset." He warned against a state of "one-way transparency," a scenario in which the United States and its allies possess advanced vulnerability-detection tools while the rest of the world remains exposed, creating an asymmetrical digital security environment.
Simultaneously, in Japan, the startup Sakana AI introduced Fugu, a frontier AI model named after the Japanese blowfish. Sakana AI, which was founded in 2023 by former Google researchers David Ha and Llion Jones alongside former Mercari executive Ren Ito, has quickly become a cornerstone of Japan’s domestic AI strategy. The company explicitly stated that Fugu "stands shoulder-to-shoulder" with Anthropic’s Fable 5 and Mythos Preview. Unlike standard large language models, Fugu is optimized as an "orchestrator," capable of managing and directing other AI agents and models through their respective APIs. This architectural choice is intended to provide Japanese businesses and government agencies with a robust, flexible infrastructure that is not dependent on a single, potentially volatile foreign provider.
Chronology of the 2026 AI Export Crisis
The current upheaval in the AI market can be traced back to a series of rapid escalations in late May and early June 2026. The following timeline outlines the key events leading to the current divergence in the global AI sector:
- May 28, 2026: Anthropic announces that its run-rate revenue has exceeded $47 billion, signaling the massive commercial success of its frontier models. The company nears a $1 trillion valuation ahead of a highly anticipated initial public offering (IPO).
- June 10, 2026: The Trump administration issues an executive order effectively banning the export of Anthropic’s Mythos and Fable 5 models to non-U.S. entities. The order cites "unacceptable risks to national security," specifically regarding the models’ ability to automate the discovery of zero-day vulnerabilities.
- June 17, 2026: At the G7 summit in Evian, France, world leaders engage in heated debates regarding AI access. Japan’s representatives, including Sakana AI co-founder Ren Ito, advocate for "AI sovereignty" and warn that hoarding technology could alienate the United States’ closest allies.
- June 22, 2026: Sakana AI launches Fugu in Tokyo, marketing it as a "frontier capability without the risk of export controls."
- June 24, 2026: 360 Security Group unveils Tulongfeng and Yitianzhen in Beijing, explicitly positioning the tools as direct competitors to the restricted American models.
Supporting Data and Market Impact
The financial and operational scale of the companies involved underscores the high stakes of this technological schism. Anthropic’s explosive growth—climbing to a $47 billion revenue run-rate by mid-2026—was driven largely by enterprise contracts. A significant portion of this growth was expected to come from the Asian market, where digital transformation initiatives are moving at a rapid pace. By restricting access to Mythos and Fable 5, the U.S. government has effectively handed a multi-billion-dollar market opportunity to domestic players in the Asia-Pacific region.
Sakana AI, while smaller than the American incumbents, has shown remarkable capital efficiency and growth. In late 2025, the company raised $135 million in a Series B round, bringing its valuation to $2.65 billion. The company’s focus on "small data" optimization and Japanese linguistic nuance has made it an attractive partner for local government agencies. The launch of Fugu is expected to further solidify its position; the model’s ability to act as an orchestrator addresses a critical vulnerability for enterprises: the risk of "overnight" loss of access to U.S.-hosted APIs.
In China, the 360 Security Group’s pivot toward AI-driven security comes at a time when the nation is doubling down on self-reliance in the semiconductor and software sectors. While 360 did not provide specific revenue figures for its new AI suite, the strategic value of Tulongfeng cannot be overstated. If the tool proves as effective as Mythos in identifying software flaws, it could neutralize the perceived advantage the U.S. sought to maintain through export bans.
Official Responses and Strategic Perspectives
The reactions from the companies and officials involved reveal a complex mix of diplomatic caution and strategic opportunism. A spokesperson for Sakana AI described the timing of the Fugu release as "entirely coincidental," noting that the research behind the model had been presented at the International Conference on Learning Representations (ICLR) earlier in the year. However, the spokesperson acknowledged that the current geopolitical climate has brought "more attention than expected" to their product.
Ren Ito, co-founder of Sakana AI, has been vocal about the dangers of technology hoarding. In a recent op-ed and during his appearance at the G7 summit, Ito argued that the United States’ first priority should be to preserve access for its allies. "AI should not become a technology that is hoarded; it should be one that is developed together," Ito stated. He cautioned that while U.S. models remain important to Asia, the current restrictions are forcing a "hedge strategy" where nations must build their own capabilities to ensure national security and economic stability.
David Ha, CEO of Sakana AI, echoed these sentiments on social media, describing "collective intelligence" as the only practical defense against the concentration of power in the hands of a few U.S. providers. "Access to top models can disappear overnight," Ha wrote, emphasizing that relying on a single foreign provider for national infrastructure is a risk that Japanese firms can no longer ignore.
In China, the rhetoric has been more assertive. Zhou Hongyi’s framing of AI as a strategic asset suggests that Beijing views the U.S. export ban not just as a commercial hurdle, but as a direct challenge to its national security. By developing Tulongfeng, 360 Security Group aims to ensure that China maintains parity in the "cyber arms race," regardless of American trade policy.
Broader Implications for Global AI Governance
The emergence of Tulongfeng and Fugu marks a turning point in the global AI race. For years, the prevailing assumption was that U.S.-based labs like OpenAI and Anthropic would maintain a multi-year lead that would force the rest of the world to remain dependent on American infrastructure. The events of June 2026 suggest that this lead may be more fragile than previously thought.
The "filling of the gap" by Asian companies suggests several long-term implications:
- Fragmentation of the AI Ecosystem: The world may be moving toward a bifurcated AI landscape, similar to the "Splinternet." One ecosystem will be built on U.S.-controlled models and values, while another—comprising China, Japan, and potentially other regional powers—will rely on locally developed, sovereign AI.
- Accelerated Local Innovation: Export bans often act as a catalyst for domestic industries. By cutting off access to the best tools, the U.S. has provided a powerful incentive for Asian governments to fund and prioritize local AI development, potentially shortening the time it takes for these companies to reach "frontier" status.
- Trust and Reliability Issues: For enterprise customers in Asia, the U.S. export ban has introduced a new category of "political risk." Even if the ban is eventually lifted, companies may be hesitant to build their core business logic on American models that could be restricted again at the whim of a future administration. Local alternatives, even if slightly less capable, offer a level of stability and cultural alignment that U.S. models cannot match.
- Strategic Vulnerability: The U.S. policy of "hoarding" cybersecurity AI could backfire if it leads to a lack of global transparency. As Zhou Hongyi noted, a world where only one side can see vulnerabilities is a world where the other side is incentivized to find and exploit them in secret, potentially making the global digital landscape less secure for everyone.
As the dust settles on the launches of Tulongfeng and Fugu, the international community is left to contemplate a new reality. The era of undisputed U.S. hegemony in frontier AI appears to be closing, replaced by a more competitive and fragmented global market where "sovereignty" is as important as "capability." Whether this leads to a more robust global intelligence or a more dangerous digital divide remains to be seen.
