Tokyo-based Sakana AI's Fugu and China-based 360's cybersecurity model Tulongfeng claim to rival Anthropic's banned Mythos and Fable 5 amid the US export ban
On Wednesday, Chinese cybersecurity firm 360 reportedly unveiled Tulongfeng, an AI tool it says can go head-to-head with Anthropic's Mythos.
Context & Ripple Effects
Related coverage first positioned 360’s Tulongfeng alongside Yitianzhen as tools for automating cyber defense. It now sits in a wider contest over substitutes for Anthropic systems that are unavailable under the US export ban.
Coverage of Anthropic’s models has focused on red-team testing and additional security-risk safeguards. That makes the new Japanese and Chinese claims relevant not only as model-performance comparisons, but as attempts to provide deployable alternatives in markets where access is constrained.
First-order effects
- Sakana AI and 360 gain a clearer opportunity to market Fugu and Tulongfeng to users unable to obtain Mythos or Fable 5; the reported claims remain vendor assertions rather than independently established parity.
- Anthropic’s restricted availability immediately leaves prospective users in affected markets to evaluate locally available model options, including tools positioned for cybersecurity use.
Second-order effects
- The claims raise pressure on rival regional AI providers to demonstrate both capability and operational security, especially where cyber-defense automation is the target use case.
- Buyers may place more weight on testing, safeguards, and integration with existing security workflows than on headline model comparisons, given the surrounding attention to Anthropic’s red-team and risk-mitigation processes.
Third-order effects
- If export restrictions persist, AI competition may fragment into regional stacks: locally available models can win distribution and enterprise integrations even without universally accepted evidence that they match frontier systems.
- The combination of cyber-defense automation and model-access controls could make security assurance a more important differentiator in AI procurement, though the corpus does not establish how customers will validate competing claims.
The trend: Export controls are creating room for regionally developed AI alternatives, while pushing model vendors to compete on deployability, security controls, and evidence of real-world capability.