Skip to content
News Model release

Sakana AI's Fugu and China's Tulongfeng fill the frontier cybersecurity AI gap left by Anthropic's export ban

· by Pondero Newsdesk

The short version

Within days of each other, Tokyo's Sakana AI and Beijing's 360 Security launched frontier-capable cybersecurity models targeting the void left by US export restrictions on Anthropic's Fable 5 and Mythos.

Sakana AI's Fugu and China's Tulongfeng fill the frontier cybersecurity AI gap left by Anthropic's export ban

Two Asian AI labs shipped competing frontier cybersecurity models in the same week, each explicitly aimed at organizations no longer able to access Anthropic's Fable 5 and Mythos. One company says the timing was accidental. The other does not.

What happened

Sakana AI, a Tokyo startup founded in 2023 by former Google researchers Ren Ito, Llion Jones, and David Ha, released Fugu Ultra on June 22, 2026. Per Sakana's announcement, Fugu Ultra "stands shoulder-to-shoulder with leading models like Anthropic's Fable 5 and Mythos Preview across the industry's most rigorous engineering, scientific, and reasoning benchmarks." The model is built as an orchestrator: rather than a standalone monolith, it routes tasks across a pool of underlying models through a single API, routing around any single provider's restrictions automatically.

Two days later, on June 24, China's 360 Security Technology took the stage at the ISC.AI 2026 cybersecurity conference in Beijing and unveiled a pair of tools it branded collectively as "Yitian Tulong." Tulongfeng is the vulnerability-detection model, and Yitianzhen is an automated cyber-defense and incident-response system. Founder Zhou Hongyi called Tulongfeng the "Chinese Mythos," and per TechRadar, claimed the tool found 3,432 flaws with 105 confirmed by the Chinese government. Zhou also acknowledged a 20 to 30 percent capability gap versus leading US models. Those figures are Zhou's own claims and have not been independently benchmarked.

Sakana told TechCrunch the Fugu release was "entirely coincidental" with the export ban, noting that the underlying research appeared at ICLR this spring. Sakana's website, though, actively markets "frontier capability without the risk of export controls." 360 Security made no coincidence claim. Zhou described vulnerability-finding AI as a "national strategic asset" and cited the risk of "one-way transparency," where some actors hold advanced capabilities while others do not, per TechCrunch.

Why it matters

The US export restriction on Fable 5 and Mythos took effect roughly two weeks before both launches per Anthropic's own announcement, creating an immediate procurement gap for enterprises and governments outside the permitted access list. Fugu and Tulongfeng are the first named products to ship directly into that opening.

The commercial stakes are real. Sakana is positioning Fugu at Japanese businesses and government agencies that want to cut exposure to export-control risk. 360 is calling its tools a national security priority. Neither product has undergone third-party review, so "comparable to Mythos" and "3,432 flaws found" remain vendor-attributed claims until independent research confirms them.

For AI tool buyers, the practical signal is this: the Fable 5 access vacuum is already being filled. If the export restriction extends beyond weeks, organizations locked out of Mythos will have tested alternatives available, with local-language optimization and no single-vendor dependency risk baked in. Sakana's multi-model orchestration architecture is specifically designed to route around any future access disruption, not just the current one.

Context

The broader pattern TechCrunch identified is that at least two companies in two different countries stepped into the same gap within days of each other. Whether that represents coordinated timing or market opportunity responding to a clear signal is a matter of interpretation. Anthropic's run-rate revenue crossed $47 billion in May 2026 per TechCrunch's earlier reporting, with Asian enterprise exposure unknown publicly. Local alternatives that understand regional language and regulatory context now exist for the first time at a frontier capability tier.

Sakana co-founder Ren Ito has separately urged the US government to preserve model access for close allies, arguing at the G7 summit in Evian last week that "AI should not become a technology that is hoarded." That context suggests Sakana views Fugu as a hedge rather than a permanent break from US AI providers.

What to watch next

Three milestones are worth tracking. First, independent benchmark results comparing Fugu, Tulongfeng, and Fable 5 once the export restriction lifts or a third-party research group gains access. Second, whether other national AI programs in the EU, India, or South Korea announce comparable capability releases in the coming weeks. Third, any sign of Sakana or 360 signing government contracts that would formalize the transition away from US providers for sovereign AI infrastructure.

Sources