The global artificial intelligence landscape shifted significantly this week as major technology firms in China and Japan unveiled high-performance AI models designed to compete directly with restricted American technology. Chinese cybersecurity giant 360 Security Technology and Tokyo-based startup Sakana AI both announced the release of new frontier models—Tulongfeng and Fugu, respectively—positioning them as alternatives to Anthropic’s Mythos, a top-tier cybersecurity AI currently under strict US export bans. These developments signal a growing trend of "technological sovereignty" in Asia, as regional players move to fill the vacuum left by the Trump administration’s recent restrictions on advanced AI exports.
The Asian Response to US Export Restrictions
On Wednesday, the Beijing-based 360 Security Technology unveiled Tulongfeng, an AI tool specifically engineered for high-stakes cybersecurity applications. According to company statements, Tulongfeng is designed to match the capabilities of Anthropic’s Mythos, a model so potent in its ability to detect and exploit software vulnerabilities that the US government has restricted its access to American citizens only. Alongside Tulongfeng, 360 also introduced Yitianzhen, a secondary tool focused on automating cyber defense and incident response.
Simultaneously, in Tokyo, Sakana AI launched Fugu, a model named after the Japanese blowfish. Fugu is marketed as a "frontier" AI model that stands on equal footing with Anthropic’s Fable 5 and the Mythos Preview. Unlike traditional large language models (LLMs), Fugu is designed as an "orchestration model," capable of managing and directing other AI agents through various APIs. This release comes just two weeks after the US government formally implemented a ban preventing Anthropic from offering Mythos and Fable 5 to international customers, citing national security concerns regarding the models’ dual-use capabilities in cyber warfare.
Chronology of the 2026 AI Export Crisis
The current friction in the global AI market is the result of a rapid escalation in regulatory and technological milestones that occurred throughout the first half of 2026:
- May 2026: Anthropic announces a historic growth milestone, reporting that its run-rate revenue crossed $47 billion. The firm nears a $1 trillion valuation ahead of a highly anticipated IPO, fueled largely by the success of its Claude, Fable, and Mythos model series.
- Early June 2026: The Trump administration issues an emergency export control order. The Department of Commerce classifies Mythos and its restricted variant, Fable 5, as "strategic cybersecurity assets." The order effectively bars any non-US entity from accessing the models’ full capabilities.
- June 15-17, 2026: At the G7 summit in Evian, France, AI access and export controls dominate the diplomatic agenda. Representatives from Japan and other allied nations express concern that a blanket ban could stifle global innovation and force allies to seek alternatives.
- June 22, 2026: Sakana AI launches Fugu, explicitly advertising the model’s ability to deliver frontier-level performance "without the risk of export controls."
- June 24, 2026: 360 Security Technology unveils Tulongfeng and Yitianzhen, framing the release as a necessary step for Chinese national security.
Sakana AI and the Strategy of Collective Intelligence
Sakana AI, 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 AI ecosystem. The company recently raised $135 million in a Series B funding round, reaching a valuation of $2.65 billion. While the timing of the Fugu release coincides perfectly with the US ban, Sakana officials maintain that the development was a long-term research goal.
"Sakana Fugu is something we have been building since last year," a company spokesperson stated. "The research behind it was presented at the International Conference on Learning Representations (ICLR) this spring. The timing simply happened to coincide with a moment that brought it more attention than we expected."
However, the strategic positioning of Fugu is clear. David Ha, Sakana’s CEO, argued on social media that relying on a single foreign provider for critical national infrastructure is a systemic risk. Ha posited that "orchestration models" represent the next frontier of AI, where the goal is not to build a single massive model, but to create a "collective intelligence" that can hedge against the concentration of power in any one country or company. By enabling agents to switch between different models via APIs, Fugu allows Japanese enterprises to remain resilient even if access to specific US-based models is revoked overnight.
360 Security and the Concept of One-Way Transparency
In China, the launch of Tulongfeng by 360 Security Technology carries even heavier geopolitical weight. Zhou Hongyi, the founder of 360, has been vocal about the dangers of what he calls "one-way transparency." This refers to a scenario where the United States possesses advanced AI tools to discover vulnerabilities in global software while denying those same defensive tools to other nations.
Zhou described vulnerability-finding AI as a "national strategic asset." From Beijing’s perspective, the US export ban on Anthropic’s Mythos is not merely a commercial restriction but a tactical move to maintain a "cyber-asymmetry." By developing Tulongfeng, 360 aims to ensure that Chinese infrastructure remains protected by domestic tools that are not subject to the whims of American foreign policy. 360’s new suite of tools is designed to automate the detection of zero-day vulnerabilities, a task that previously required thousands of man-hours from highly skilled security researchers.
Economic Implications and Market Realignment
The economic stakes of these export controls are immense. Anthropic’s $47 billion run-rate revenue in May 2026 reflected a massive global demand for high-end AI. A significant portion of that revenue was expected to come from the Asia-Pacific region, where digital transformation is accelerating in both the public and private sectors.
By cutting off access to Mythos and Fable 5, the US government has inadvertently created a market opportunity for regional competitors. While Sakana AI’s spokesperson noted that "U.S. models remain important to Asia," the firm’s marketing strategy suggests a shift in buyer behavior. Japanese businesses and government agencies are increasingly looking to reduce their exposure to tightening export controls.
Data suggests that local models hold a distinct advantage in specific markets:
- Cultural and Linguistic Nuance: Models like Fugu are optimized for the Japanese language and cultural context, often outperforming generalized Western models in local business logic.
- Data Sovereignty: Local models allow firms to keep sensitive data within national borders, complying with increasingly strict regional privacy laws.
- Stability of Supply: As David Ha noted, "access to top models can disappear overnight." For critical infrastructure, the reliability of a local partner outweighs the raw performance edge of a sanctioned foreign product.
Analysis: A Permanent Realignment?
The emergence of Tulongfeng and Fugu raises the question of whether the global AI market is entering a permanent state of fragmentation. During the G7 summit in Evian, Sakana co-founder Ren Ito argued in an op-ed that the US federal government’s first priority should be to "preserve access" for its closest allies. He warned that "AI should not become a technology that is hoarded; it should be one that is developed together."
Despite these warnings, the trend toward localized AI clusters seems to be accelerating. If US companies like Anthropic and OpenAI are prevented from serving international markets, they risk losing the network effects and data feedback loops that have maintained their dominance. Conversely, if firms in Tokyo and Beijing can prove that their models are "good enough" for enterprise and security needs, the trust in US-based "AI-as-a-Service" may never fully recover, even if the bans are eventually lifted.
The current situation mirrors previous shifts in the telecommunications and semiconductor industries, where export restrictions led to the rise of robust domestic competitors in Asia. As Tulongfeng and Fugu begin their deployment, the AI industry may be witnessing the birth of a multipolar technological world, where the "frontier" is no longer defined solely by Silicon Valley.
