What are Sakana AI's Fugu and 360 Security's Tulongfeng?

Sakana AI's Fugu is a new AI tool focused on orchestrating multiple models to achieve performance benchmarks comparable to Anthropic's Fable 5. 360 Security's Tulongfeng is a vulnerability discovery tool designed to rival the cybersecurity capabilities of Anthropic's Mythos. Both are Asian-developed alternatives to the restricted Western models.

Why are these new launches significant?

These launches are significant because they directly address a market gap created by a major AI developer's export restrictions. They demonstrate the capability of Asian tech firms to rapidly develop high-performance AI alternatives, potentially reshaping regional AI market dynamics and contributing to a more fragmented, yet resilient, global AI ecosystem.

Will these new tools fully replace Anthropic's models?

While Fugu and Tulongfeng are positioned as direct alternatives, whether they fully replace Anthropic's models depends on their actual performance, scalability, and market adoption. Their advantage lies in local language support and immediate availability in markets where Anthropic's models are restricted, but long-term success will require sustained innovation and trust.

Image: courtesy of Thenextweb

techJune 28, 2026By Veridact EditorialUpdated Jun 28

Anthropic's Export Ban Opens Door for Asian AI Challengers in Tokyo and Beijing

A Tokyo-based startup, Sakana AI, and a Beijing security firm, 360 Security, have launched new artificial intelligence tools, Fugu and Tulongfeng, respectively. These launches come weeks after Anthropic's powerful AI models, Mythos and Fable 5, became subject to an export ban, creating a significant market vacuum, particularly in Asia. The new offerings are designed to provide alternatives, with Fugu focusing on AI orchestration and Tulongfeng on vulnerability discovery, directly challenging the capabilities of Anthropic's restricted models.

Outlook

The immediate aftermath of these launches will likely see a focused push by Sakana AI and 360 Security to capture market share in Asia. Enterprises and developers in the region, previously interested in Anthropic's advanced models, are now looking for viable alternatives that are not subject to similar geopolitical restrictions. This situation creates a fertile ground for local players who can offer solutions tailored to regional languages, regulatory environments, and specific business needs.

We can expect an intense period of competitive positioning, with these new entrants showcasing their models' performance against established benchmarks and emphasizing their local advantages. The market will be watching closely to see how quickly these tools can integrate into existing workflows and whether they can indeed deliver on their promises to rival the capabilities of their Western counterparts. Anthropic, while maintaining its partner-focused strategy for Mythos, will likely monitor the market penetration of these new Asian competitors.

Background

The current market shift stems directly from Anthropic's decision to implement an export ban on its more powerful AI models, specifically Claude Mythos and Fable 5. Claude Mythos, previewed on April 8, 2026, is known for its strong cybersecurity capabilities, particularly in detecting software vulnerabilities. However, its announcement on April 10, 2026, also sparked a debate about the safety implications of releasing such powerful open-weight AI models, with critics expressing concerns about digital infrastructure preparedness.

Anthropic confirmed that Mythos will not be made publicly available, instead limiting access to a select group of partners such as Microsoft and Amazon. This restriction effectively created a void for businesses and developers outside these privileged partnerships, especially those in regions like Asia, who were seeking cutting-edge AI for tasks like complex orchestration and advanced cybersecurity.

Into this void step Sakana AI and 360 Security. Sakana AI's offering, Fugu, is designed for AI orchestration, meaning it can manage and coordinate multiple AI models to work together. The company claims Fugu can match Fable 5 benchmarks without needing to train a 'frontier model' from scratch, suggesting an efficient and potentially cost-effective approach. Meanwhile, Beijing's 360 Security has launched Tulongfeng, a vulnerability discovery tool that the firm states rivals Anthropic's Mythos in its capabilities. These new tools are explicitly positioned as regional alternatives, addressing the demand for powerful AI that understands local languages and nuances, a significant advantage in diverse Asian markets.

Precedents

History is replete with examples of market gaps created by export controls, trade restrictions, or geopolitical tensions leading to the rise of local champions. During the Cold War, restrictions on technology transfers often spurred domestic innovation in both blocs. More recently, U.S. sanctions against specific Chinese technology firms have accelerated China's drive for self-sufficiency in critical areas like semiconductors, leading to significant investments and advancements in local manufacturing and R&D.

In the software sector, similar dynamics have played out. When certain foreign software or services become unavailable or politically sensitive, local alternatives often emerge quickly, sometimes even surpassing the original offerings due to their tailored approach to local user needs, language, and regulatory frameworks. The internet industry in China, for instance, developed its own robust ecosystem of social media, e-commerce, and search platforms largely independent of Western giants, partly due to regulatory barriers and partly due to a deep understanding of local consumer behavior.

This pattern suggests that while export bans might limit the reach of a specific technology, they rarely stifle innovation entirely. Instead, they often redirect talent and capital towards creating indigenous solutions, fostering a more diversified and regionally focused technological landscape. The current situation with Anthropic's models and the rapid response from Sakana AI and 360 Security aligns squarely with this historical precedent.

The emergence of Fugu and Tulongfeng in response to Anthropic's export ban carries several significant implications. First, it highlights the immediate and tangible impact of geopolitical decisions on the global technology market. A regulatory move by one nation or company can instantly reshape competitive dynamics thousands of miles away, creating both challenges and opportunities.

For businesses and developers in Asia, these new tools offer crucial access to advanced AI capabilities that might otherwise be unavailable. This could prevent a 'digital divide' where access to cutting-edge AI is restricted to a select few, ensuring broader participation in the AI economy. It also speaks to the growing maturity and ambition of Asian AI ecosystems, demonstrating their capacity to develop and deploy sophisticated models rapidly.

From a broader perspective, this development could accelerate the fragmentation of the global AI market. As more regional players develop their own powerful models, optimized for local contexts and potentially adhering to different ethical or regulatory standards, the vision of a single, interconnected global AI infrastructure may give way to more localized, interconnected ecosystems. This could foster greater resilience and diversity in AI development but also introduce complexities in interoperability and global standards.

Ultimately, the real stakes here are about technological sovereignty and competitive advantage. Nations and companies are increasingly aware that control over foundational AI models translates into economic and strategic power. The rapid response from Tokyo and Beijing is a clear signal that they intend to compete at the highest levels, irrespective of export controls from Western firms.

Scenarios

Analysis

One possible outcome is that Sakana AI's Fugu and 360 Security's Tulongfeng gain significant traction within Asian markets. Their focus on local language and cultural nuances, combined with the absence of Anthropic's top-tier models, creates a strong incentive for regional businesses to adopt these alternatives. This could establish them as dominant players in their respective niches across Asia, potentially leading to further investment and expansion for both companies.

Another scenario could see Anthropic reassessing its export strategy. If the market gap continues to be filled effectively by non-Western competitors, Anthropic might face pressure from its partners or investors to find ways to expand its global reach, perhaps through licensing agreements or modified models that meet export requirements. However, any such shift would likely be complex, given the underlying geopolitical and safety concerns that led to the ban in the first place.

Alternatively, the rapid development and deployment of these Asian alternatives could trigger a broader 'AI arms race' where other regions or nations double down on developing their own sovereign AI capabilities. This could lead to a proliferation of national AI champions, each vying for technological leadership and potentially operating under distinct regulatory frameworks, further fragmenting the global AI landscape and complicating future international collaborations.

Finally, the performance and security of these new models will be under intense scrutiny. If Fugu and Tulongfeng consistently deliver on their performance claims and demonstrate robust security, it could validate the strategy of developing powerful AI outside the established Western tech giants. Conversely, any significant shortcomings could slow their adoption and allow other emerging players to challenge their position.

Timeline

2026-04-08

Claude Mythos Preview Launch

Anthropic launched a preview of its Claude Mythos model, showcasing its advanced capabilities, particularly in software vulnerability detection.

2026-04-10

Open-Weight Safety Debate Ignites

The announcement of Anthropic's Claude Mythos model sparked renewed arguments regarding the safety and responsible release of open-weight AI models, with critics citing unprepared digital infrastructure.

2026-06-03

Anthropic Export Ban Takes Effect (Inferred)

Weeks prior to the new launches, Anthropic's export order on its powerful models like Mythos and Fable 5 took effect, limiting their availability to a select group of partners like Microsoft and Amazon.

2026-06-27

Sakana AI and 360 Security Launch New Tools

Tokyo's Sakana AI launched Fugu for AI orchestration, and Beijing's 360 Security launched Tulongfeng for vulnerability discovery, both positioned as alternatives to Anthropic's restricted models.

Frequently Asked Questions

Anthropic's export ban refers to the company's decision to restrict the broad public release of its most powerful AI models, such as Claude Mythos and Fable 5. Instead, access is limited to a select group of partners like Microsoft and Amazon, effectively creating a market void for other businesses and developers, particularly in regions like Asia.

Discussion

Be the first to share your thoughts.