Z.ai reported that Chen and his team identified 23 security vulnerabilities in Microsoft software using the company’s GLM artificial‑intelligence models, and Microsoft confirmed the findings. This shows that AI can help uncover real‑world bugs in widely used products.

Z.ai says its newest GLM‑5.3 model finds bugs much faster than earlier versions, marking a clear step forward for AI‑driven security research. The claim was made in August, before the model was evaluated by external experts.

An assessment by the U.S. AI firm Anthropic later found that GLM‑5.3 demonstrates advanced capabilities in identifying and exploiting cybersecurity weaknesses. The independent review supports Z.ai’s claim that the model improves vulnerability discovery.

Z.ai was founded in 2019 by researchers linked to China’s Tsinghua University and has recently been valued at about US$32 billion after a large fundraising round. Chen’s move highlights how AI startups are becoming major players in the fight against software threats.