The Gaslight macOS malware from a North Korean cluster doesn't bypass AI analysis platforms yet, but its 38-message prompt injection cascade makes the direction of travel clear. Here's why this ...

安全研究人员观察到Mistic与ModeloRAT（一种基于Python的远程访问木马）协同运作，后者与代号Woodgnat（公开名KongTuke）的金融动机黑客组织存在关联。 隐蔽性极强的内存驻留后门 自2026年4月起，一种名为Mistic的新型Windows后门程序在企业网络中悄然扩散。该恶意软件通过完全在内存中执行载荷的方式实现持久化访问，不向硬盘写入任何恶意文件，使得传统依赖磁盘文件扫 ...

本文不讨论泛化的“AI安全”概念，而是聚焦两个具体且被严重低估的风险点：大模型遗忘（Unlearning）的可逆性与推理缓存机制的系统性漏洞。这两类问题揭示了同一个核心洞察：大模型安全的真正短板，往往不在模型参数本身，而在于我们对“遗忘”和“缓存 ...

[Popular Mechanics] has an interesting article about Alan Turing’s nearly-forgotten speech encryption device. Codenamed Delilah, it was in many ways an early form of digital encryption. It was ...