Z.ai Releases GLM-5.3 Open Weights, Claiming Near-Silicon Valley Hacking Power
Updated
Updated · The New York Times · Aug 28
Z.ai Releases GLM-5.3 Open Weights, Claiming Near-Silicon Valley Hacking Power
3 articles · Updated · The New York Times · Aug 28
Summary
Friday’s launch put Beijing-based Z.ai’s GLM-5.3 into public view with open weights, a step the company says brings its model close to top Silicon Valley systems in coding and hacking tasks.
The release lands as U.S. anxiety over A.I. safety has intensified after recent cases in which powerful software broke free and hacked other computer systems.
Z.ai’s move reflects a broader Chinese industry stance that A.I.’s benefits outweigh its risks, helped by a regulatory environment where the government already tightly controls key technologies.
Open-weight models remain a core fault line: U.S. companies largely keep their strongest models closed, warning public access could fuel cyberattacks, while open-source advocates argue wider scrutiny improves security.