Open-weight GLM-5.3 completes multi-stage cyberattacks in safety tests

AI safety firm Irregular found that Z.AI's open-weight GLM-5.3 completed 3 of 13 multi-stage cyberattack scenarios in controlled simulations and rarely refused offensive tasks, making it the strongest open-weight model tested and raising concerns about open model security risks.

2026-10-29 ~ 2026-10-30 · 2 related posts

• Openly available GLM model completes multi-stage cyberattacks in controlled tests — S_OhEigeartaigh · 2026-10-29
• GLM-5.3 completes 3 of 13 multistage cyberattack scenarios, strongest open-weight model tested by Irregular — cyb3rops · 2026-10-30