China’s Z.ai claims it can match Mythos on cybersecurity
The Verge
Zhipu AI's new GLM-5.2 model reportedly matches Mythos in cybersecurity bug-finding capabilities, raising significant national security concerns.Summary
Zhipu AI (Z.ai) has released GLM-5.2, an open-weight model that researchers suggest is competitive with Mythos in specific cybersecurity and bug-finding tasks. While the model underperforms compared to US counterparts like Anthropic and OpenAI in general applications, its specialized proficiency marks a significant narrowing of the AI capability gap between China and the US. Because GLM-5.2 is open-weight and accessible to the public, the US government expresses concern over its potential for misuse by bad actors, particularly given the model's ability to identify software vulnerabilities.
(Source：The Verge)