The global artificial intelligence ecosystem is experiencing a profound realignment following the Trump administration’s decision to restrict access to advanced U.S. AI models. Within weeks of the United States barring non-Americans from utilizing Anthropic’s high-powered cybersecurity-focused models—namely Mythos and its more restricted counterpart, Fable 5—major technological players in Asia have accelerated the deployment of indigenous alternatives.
The most
notable developments arrived from Tokyo and Beijing within days of each other. Tokyo-based AI startup Sakana AI launched "Fugu," a frontier model designed specifically for agent orchestration and multi-model API access. Simultaneously, Chinese cybersecurity powerhouse 360 unveiled "Tulongfeng," an AI tool engineered to compete directly with Anthropic’s restricted technologies by automatically discovering software vulnerabilities.
These rapid product rollouts underscore a growing anxiety among international governments and enterprises regarding technological sovereignty. As export controls tighten, the reliance on single-provider U.S. infrastructure is increasingly viewed as a structural vulnerability rather than a mere commercial convenience.
Chronology of the Export Ban and Asian Responses
The current geopolitical friction traces back to a decisive regulatory shift enacted by the United States government. Approximately two weeks prior to the recent Asian product launches, the Trump administration issued an executive order restricting global access to Anthropic’s flagship cybersecurity model, Mythos, and the Fable 5 preview. U.S. officials defended the move as a necessary national security safeguard, citing the immense, dual-use capabilities of AI models capable of autonomously identifying critical software vulnerabilities at unprecedented speeds.
However, the ripple effects were felt immediately across international markets. On Monday, Sakana AI introduced Fugu—named after the Japanese word for blowfish—claiming it stands shoulder-to-shoulder with leading U.S. models. While Sakana leadership insisted that the launch timing was entirely coincidental and that the underlying research had been presented earlier this spring at the International Conference on Learning Representations (ICLR), the company quickly pivoted to capitalize on the regulatory vacuum.
By Wednesday, the technological response escalated further east. Chinese cybersecurity firm 360 officially debuted Tulongfeng, alongside a companion tool named Yitianzhen, which is built to automate cyber defense and incident response. Unlike Sakana, which framed its product as a cooperative hedge against single-provider risk, Chinese firm 360 framed its new AI capabilities through the lens of national strategic defense against what it termed "one-way transparency."
Architects of Collective Intelligence: Sakana AI’s Strategy
Sakana AI, co-founded in 2023 by former Google researchers David Ha and Llion Jones alongside former Mercari and Stability AI executive Ren Ito, has carved a distinct niche in the generative AI market. The Tokyo-based startup specializes in cost-effective models optimized for smaller datasets, with a particular emphasis on Japanese language, culture, and localized nuance.
Despite timing its release during a period of intense U.S. export scrutiny, Sakana leadership has been careful to temper narratives of a complete permanent decoupling from American technology. During the G7 summit in Evian, Ren Ito emphasized that U.S. models remain vital to the Asian enterprise ecosystem, characterizing the current climate as a temporary adaptation rather than a permanent fracture.
This sentiment was echoed in an op-ed published by Ito in Project Syndicate, where he urged Western policymakers to prioritize preserving access for close allies. Ito argued that artificial intelligence should function as a shared, collaborative technology rather than a hoarded strategic asset.
Nevertheless, Sakana’s CEO David Ha stressed that Fugu represents an essential evolutionary step for enterprise architecture. By positioning Fugu as an orchestration model capable of routing tasks across multiple APIs, Sakana is addressing a newly awakened corporate fear: that access to foundational infrastructure can be revoked overnight. "Access to top models can disappear overnight," Ha noted via social media commentary. "Collective intelligence is the practical hedge against this concentration of power."
Beijing’s Counter-Offensive: 360 and National Cyber Sovereignty
While Tokyo’s approach leans toward multi-model orchestration and hedging, Beijing’s response under 360 reflects a more direct challenge to Western AI dominance. Founder Zhou Hongyi publicly designated vulnerability-finding artificial intelligence as a critical national strategic asset.
In statements reported by international media outlets, Zhou raised concerns regarding "one-way transparency"—a scenario where select geopolitical actors monopolize advanced vulnerability-detection capabilities while leaving rival nations exposed. Tulongfeng and Yitianzhen are designed precisely to neutralize this imbalance by automating both the offensive discovery of software flaws and the defensive incident response protocols required to patch them.
The introduction of these tools highlights a broader bifurcation in global cyberspace. For years, Chinese enterprises heavily relied on open-source weights originating from Western labs or sought workarounds to access closed commercial APIs. The latest U.S. export restrictions, however, have provided local champions like 360 with both the political mandate and the market motivation to accelerate domestic capability self-sufficiency.
Market Implications for Anthropic and the Broader AI Industry
The timing of these developments comes at a critical juncture for Anthropic. Prior to the export ban, the San Francisco-based AI lab was scaling at a historic pace, with reports in May indicating that its run-rate revenue had crossed $47 billion ahead of a widely anticipated initial public offering.
While the exact percentage of Anthropic’s revenue derived from Asian enterprise customers remains proprietary, the sudden foreclosure of these markets introduces fresh variables into the company’s long-term financial projections. More importantly, market analysts point out a structural risk for U.S. tech firms: once foreign enterprises are forced to adopt domestic alternatives due to regulatory compliance, winning back those market shares becomes exceptionally difficult.
Local alternatives are not merely replicating U.S. benchmarks; they are inherently better attuned to regional regulatory frameworks, linguistic nuances, and localized operational requirements. Even if Washington were to lift the export ban on Mythos and Fable 5 in the future, the foundational groundwork laid by startups like Sakana AI and established giants like 360 ensures that the monopoly once held by Silicon Valley on frontier intelligence has been permanently fractured.
As governments from Tokyo to Beijing recalibrate their digital infrastructure policies around the concept of AI sovereignty, the industry is transitioning away from a unipolar market dominated by a handful of U.S. labs. Instead, the future of artificial intelligence appears increasingly fragmented, multi-polar, and defined by regional resilience against geopolitical volatility.



