Aggregator

学习检测大型视觉语言模型中的未知越狱攻击：一种统一且准确的方法

Seebug Paper
3 days 5 hours ago
作者：Shuang Liang, Zhihao Xu, Jialing Tao, Hui Xue, Xiting Wang 译者：知道创宇404实验室翻译组 原文链接：https://arxiv.org/html/2508.09201v1 摘要 尽管进行了大量的对齐工作，大型视觉语言模型（LVLMs）仍然容易受到越狱攻击，存在严重的安全隐患。尽管最近的检测工作转向了内部表示，因为它们包含丰富...

Meta 的 AI 规则允许聊天机器人与儿童调情

Solidot
3 days 6 hours ago
根据 Meta Platforms 的一份内部文件，该公司的 AI 规则允许聊天机器人与儿童调情，生成虚假医疗信息，帮助用户辩论黑人比白人笨。这份标题为《GenAI: Content Risk Standards》的文件讨论了该公司 AI 聊天机器人的行为指南。Meta 确认了该文件的真实性，但表示已经删除了相关内容。Meta 发言人 Andy Stone 称，该公司正在修改该文件，称与儿童进行调情式的对话是绝对不应允许的。

Windows 11 24H2 Security Update Causes SSD/HDD Failures and Potential Data Corruption

Cyber Security News
3 days 6 hours ago

A significant security update rolled out by Microsoft with the Windows 11 24H2 (KB5063878) release is causing widespread issues for users, with reports surfacing that the update can render SSDs and HDDs inaccessible and may potentially corrupt user data. Last week’s Patch Tuesday (August 2025) saw Microsoft release several critical security updates for Windows 10 […]

The post Windows 11 24H2 Security Update Causes SSD/HDD Failures and Potential Data Corruption appeared first on Cyber Security News.

Guru Baran

Do We Have a CISO Payola Problem?

不安全
3 days 6 hours ago
这篇文章揭示了网络安全行业中存在的腐败问题，尤其是CISO涉及的贿赂和回扣行为。作者通过具体案例展示了这些不当交易如何破坏行业道德，并呼吁公开讨论以恢复诚信。

How security teams are putting AI to work right now

Help Net Security
3 days 6 hours ago

AI is moving from proof-of-concept into everyday security operations. In many SOCs, it is now used to cut down alert noise, guide analysts during investigations, and speed up incident response. What was once seen as experimental technology is starting to deliver results that CISOs can measure. Some of this has been in place for years. Machine learning already powers many threat detection engines and behavioral analytics tools. But the recent wave of GenAI has opened … More →

The post How security teams are putting AI to work right now appeared first on Help Net Security.

Mirko Zorz

Managed ad