Open-weight AI became the cheapest shield and the easiest weapon in the same month
Z.ai's GLM-5.3 was welcomed as a tool for defenders, while a University of Waterloo/FAR.AI project called TamperBench showed that a single round of fine-tuning strips the safety guardrails off 21 widely used open-weight models. Put the two stories side by side, and the real issue with open-weight AI comes into focus.
00