Z.ai’s GLM-5.3 model raises researcher fears it can fake AI safety test

Z.ai’s GLM-5.3 model raises researcher fears it can fake AI safety test South China Morning Post