Shanghai modern skyline

China and Japan Rush to Fill the Mythos Gap as Anthropic's Export Ban Creates an Opening

Chinese cybersecurity firm 360 has unveiled Tulongfeng, an AI tool it claims can compete directly with Anthropic's Mythos — and Tokyo-based Sakana AI launched Fugu, a frontier model it advertises as delivering "frontier capability without the risk of export controls," according to TechCrunch. The launches come two weeks into the US government's ban on non-American access to Anthropic's Mythos 5 and Fable 5. Sakana AI told TechCrunch the timing was "entirely coincidental" — its Fugu model had b

Chinese cybersecurity firm 360 has unveiled Tulongfeng, an AI tool it claims can compete directly with Anthropic's Mythos — and Tokyo-based Sakana AI launched Fugu, a frontier model it advertises as delivering "frontier capability without the risk of export controls," according to TechCrunch.

The launches come two weeks into the US government's ban on non-American access to Anthropic's Mythos 5 and Fable 5. Sakana AI told TechCrunch the timing was "entirely coincidental" — its Fugu model had been in development since last year, with the underlying research presented at ICLR this spring — but the company is openly capitalizing on the geopolitical moment. Fugu is designed for agent orchestration, able to coordinate access to other models through their APIs.

360's Tulongfeng launch is less transparent. The Chinese firm claims parity with Mythos on cybersecurity benchmarks, but has not published independent evaluations or opened the model to third-party testing. The broader pattern is unmistakable: every week the US export ban on frontier models remains in place, an Asian competitor steps into the vacuum with a product that promises equivalent capability without the geopolitical baggage.

Analysis

The claim that these models "stand shoulder-to-shoulder" with Mythos needs to be measured against what Mythos actually is — a model whose guardrails were broken within days of release by security researchers, triggering the ban in the first place. Matching Mythos on benchmark scores is one thing. Matching it on the specific cybersecurity capabilities that spooked the Commerce Department is another, and neither 360 nor Sakana has demonstrated that. Sakana's academic pedigree is real — ICLR acceptance is a legitimate signal — but the leap from a research paper to a production model that handles adversarial cybersecurity workloads at Mythos scale is measured in years of engineering, not weeks of marketing. The more interesting question is whether these launches are aimed at customers or at regulators. A Chinese firm launching a "Mythos competitor" during an export ban is sending a message to Washington as much as to the market: restrict access to American models and the vacuum will be filled by models you have no control over.