智谱 GLM-5.3 扫描开源项目，披露 2400 余漏洞
cedric_chee · x · 2026-08-14
A user reported that GLM-5.3 is scanning open-source projects and popular software at scale, publicly disclosing the vulnerabilities it finds via the Z.ai Security platform.
The disclosure ledger shows 2,436 vulnerabilities cataloged, including 1,097 critical and high-severity flaws. These vulnerabilities span an impact timeline of 45 years, with an average dormancy period of 26.6 years. Recent high-severity disclosures affect core components like the Linux kernel, WebKit, and FreeBSD.
所属事件:智谱GLM模型大规模扫描开源项目发现超2400个漏洞(2 条相关)→
「模型」频道最新
- 实测Qwen 3.8 Max：规划能力突出，擅长简化复杂架构 — RevealIndividual7567 · 2026-08-14
- 用户吐槽 Claude 安全分类器误报，质疑模型回退至 Opus 4.8 — davidmanheim · 2026-08-14
- Qwen 27B dense在RTX 3090上跑出40 tok/s，本地AI的浪漫 — danbri · 2026-08-14
- LoRA 训练难题：Qwen 图像模型无法生成全身像 — Kind-Illustrator6341 · 2026-08-14
- 智谱发布GLM-5.3，自称最强开源编码模型，性能提升50% — The Decoder · 2026-08-14
- 知名工程师上手 Grok 4.6:与 4.5 差异明显,训练见功力 — MaziyarPanahi · 2026-08-14