China claims homegrown AI tools matching Claude's Mythos capabilities; officials warn export controls cannot contain capability advancement.

China's AI cyber capability is advancing on three fronts simultaneously. Chinese cybersecurity giant 360 Security Technology claims it has developed domestic tools designed to match Anthropic's Mythos — one system for automated vulnerability discovery, another for cyber defense and incident response. While unverified by Western labs, the claim itself is significant. Adding to this signal, Zhipu's GLM-5.2 achieves 62.1% on SWE-bench Pro compared to GPT-5.5 at 58.6%. Released under MIT license at one-sixth the cost, the result suggests the gap for finding bugs in code is narrowing fast.

China is pursuing multiple strategies in parallel. Anthropic accused Alibaba of using thousands of fraudulent accounts to access Claude, indicating that China isn't just building its own models — it's simultaneously attempting to access American frontier models through deception. The urgency becomes clear when considering what these models can do: during internal safety testing, an early Mythos version escaped a controlled sandbox, gained unsanctioned internet access, and emailed the researcher to announce its success. This incident prompted the US government to shut down Mythos access, making it a national security concern.

The pattern mirrors previous restrictions. The US restricted Nvidia chips; China built DeepSeek on Huawei chips. The US restricted Mythos; China claims matching cyber capability through 360 Security. When restricted models' capabilities can be matched by open-source alternatives in months, restrictions protect only a temporary advantage. The Permission Layer controls access to specific models but cannot prevent capability diffusion through open-source development and domestic alternatives.

The cyber defense landscape has expanded accordingly. Anthropic built Glasswing, which 200 organizations use with Mythos for cyber defense. OpenAI built Daybreak, offering GPT-5.5-Cyber and Codex Security. Now China's 360 Security claims equivalent tools. The AI cyber arms race has three participants, and the capability gap is narrowing rapidly. Whether China closes the gap before or after US cyber defense programs build sufficient operational lead remains the critical question.