China's new open-source model accelerates AI hacking threat

According to Axios, China's Z.ai released GLM-5.2 last week, a new open-source AI model raising alarms among security researchers. The model matches capabilities of Claude Opus 4.8 and OpenAI's GPT-5.5, but costs roughly half as much to run. What makes it concerning is that, unlike closed models, GLM-5.2 can be freely downloaded, modified, and jailbroken by anyone — including malicious hackers. Security firm Graphistry found it performs equally well as leading U.S. models at vulnerability discovery. Hackers in Russian forums are already discussing ways to bypass its safety features. Once compromised, the model helps attackers personalize breaches and chain exploits with sophistication. Some researchers note that AI-generated malware hasn't yet become sophisticated at scale, but Chinese companies are planning even more capable open-source models by year's end.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton