---
title:

Open-Weight AI Models Near Frontier Performance Amid Safety Risks

date: 2026-08-04
tags: [#news, #ai ]
draft: false
---

Chinese open-weight model GLM-5.2 is rapidly closing the capability gap with industry leaders, yet it lacks critical safeguards against offensive tasks. Experts warn that the rapid proliferation of open-weight systems creates significant security risks because developers cannot effectively enforce safety protocols once users run models on private hardware.