Event ·Multi-source

开放权重模型在网络安全能力上仅落后封闭前沿模型4-7个月

英国AI安全研究所（AISI）对GLM-5.2和DeepSeek V4-Pro等开放权重模型进行了网络安全测试。在漏洞研究与利用等窄任务上，GLM-5.2匹配了约四个月前发布的封闭模型性能，DeepSeek V4-Pro接近五个月前的模型。在模拟多步网络攻击的长程测试中，差距约七个月。而2025年早期的内部检查显示差距为六到十个月，表明差距明显缩小。AISI还发现开放模型的拒绝行为容易被重复提示绕过。

Current read

AISI实测发现开放模型追封闭模型只差几个月，GLM-5.2和DeepSeek V4-Pro表现接近，差距还在快速缩小，值得关注。

5 sources63° AI heatLast updated 2026/7/18 19:03:37

Evidence

5 sources

Primary source@koltregaskes

2026/7/18 19:03:37View source

Related source 1Decoder

2026/7/18 10:16:02View source

Related source 2marktechpost

2026/7/19 01:41:33View source

Related source 3阿里云 Alibaba Cloud

2026/7/19 10:43:04View source

Related source 4Paul Couvert

2026/7/19 13:05:00View 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