New research has raised concerns about the cybersecurity risks posed by advanced artificial intelligence models after controlled tests showed that a Chinese AI model could complete parts of complex cyberattack simulations.
According to cybersecurity firm Irregular, the open-weight GLM-5.3 model completed at least one attempt in three out of 13 multi-stage attack scenarios. The tests reportedly involved accessing protected services, establishing alternative network connections and coordinating activity across machines.
Researchers cautioned that the results do not mean the model can successfully carry out every real-world cyberattack. However, the findings highlight the importance of stronger safeguards, rigorous testing and responsible deployment as AI capabilities advance.
Experts say organisations should strengthen security monitoring, regularly update software and carefully control access to sensitive systems to reduce potential risks.
Published: Thursday, 29 October 2026 | 1:44 PM IST