Open Source Model Rivals Proprietary Security AI
Chinese AI startup Zhipu AI has released GLM 5.2, an open weight model that security researchers say performs on par with Anthropic’s Mythos in identifying software vulnerabilities. According to a Wall Street Journal report, GLM 5.2 matches the capability of Mythos, widely considered the gold standard for AI powered cybersecurity, particularly in bug finding tasks.
The model is open source, allowing anyone to download and run it on their own hardware without relying on a cloud provider. While GLM 5.2 still trails Anthropic and OpenAI in broader reasoning tasks, it has reportedly outperformed Claude Opus 4.8 in some security evaluations. With additional prompting, researchers say it can reach Mythos level performance in vulnerability detection.
Impact on US China AI Policy
This development comes at an awkward time for US AI policy. While companies like Anthropic and OpenAI have been restricting access to their most advanced frontier models over national security concerns, Chinese labs have been releasing increasingly capable open weight alternatives. The narrowing gap in cybersecurity AI has significant implications. Cybercriminals could potentially adapt GLM 5.2 for offensive purposes, and the findings suggest China may be closing the AI capability gap faster than expected.
Elon Musk recently predicted Chinese AI labs would catch up to Anthropic’s flagship Fable 5 by Q1 2027 on benchmarks. Zhipu AI founder Tang Jie pushed back, saying it “won’t take that long,” and this WSJ report appears to give that optimism some weight.
Source: The Wall Street Journal
