Event ·Multi-source

Anthropic测试：GLM-5.3具备网络攻击能力

Anthropic Frontier Red Team团队在100项二进制利用基准测试中发现，GLM-5.3在4%的试验中实现了完整控制流劫持。Claude Mythos Preview的表现略高，达到6%。此前模型如Claude Opus 4.6和GLM-5.2在相同测试中均未成功。

Current read

Anthropic测试显示GLM-5.3首次具备网络攻击能力，比前代模型有明显提升。

11 sources58° AI heatLast updated 2026/9/29 22:20:28

Evidence

11 sources

Primary sourceSimon Willison’s Weblog

2026/9/29 22:20:28View source

Related source 1Teortaxes

2026/9/29 21:18:50View source

Related source 2宝玉

2026/9/29 23:14:25View source

Related source 3shao__meng

2026/9/30 02:07:01View source

Related source 4Decoder

2026/9/30 11:05:05View source

Related source 5arXiv: DeepSeek

2026/9/28 06:45:01View source

Related source 6IT之家

2026/9/28 23:40:42View source

Related source 7AI Will

2026/9/29 07:51:37View source

Related source 8Lxfater

2026/9/30 03:58:31View source

Related source 9orange.ai

2026/9/28 00:17:04View source

Related source 10lmarena.ai

2026/9/28 17:00:36View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.

Open the story