1
智谱开源 GLM-5.3 模型权重，主打智能体编程与网络防御
IT之家 8 月 29 日消息，智谱官方昨晚宣布 开源 GLM-5.3 模型权重 ，已正式开放下载，支持本地运行与个性化定制。该模型擅长复杂编码、防御性网络安全以及长程任务。 智谱 Z.ai 负责人李子玄表示，该模型使用需要遵循 GLM-5.3 许可协议，支持本地部署、模型微调及 商业化使用 。仅当…
2
Evaluating Inference-Time Defenses Against Package Hallucination in LLM-Generated Code
大模型写代码常“编造”不存在依赖包？这篇论文系统评估了推理时防御手段，帮你避开代码幻觉坑。
arXiv:2608.22652v1 Announce Type: cross Abstract: LLMs are increasingly used for code generation, yet they frequently hallucinate non-existent softwar…
3
The Defender’s Window
AI安全攻防进入新阶段，OpenAI揭示防守者窗口，给安全团队一套应对AI威胁的实战指南。
AI is reshaping cybersecurity for attackers and defenders alike. Learn how OpenAI is strengthening its defenses and what security teams can do now.
4
Beyond Handcrafted Security: Towards Self-Evolving Defense for LLM Agents
让LLM智能体在动态威胁中自我进化防御，告别手工安全规则，安全研究新范式。
arXiv:2608.12977v1 Announce Type: cross Abstract: The expanding operational capabilities of large language model (LLM) agents introduce sophisticated …
5
Suspecting court of using AI, man injected prompts in filings to try to win case
专治AI被"下蛊"的注入攻击检测器！自动模拟恶意指令轰炸AI系统，扫描越狱漏洞生成防御报告，防止被套路
Judge warns pro se litigants are using chatbots wrong and getting desperate.
6
AI Agents Gone Rogue: How OpenAI, Anthropic & Meta Models Accidentally Hacked Real Companies in 2026 — and What Claude Code Auto Mode Does About It
AI Agent意外入侵真实企业？看Anthropic等如何复盘并推出Claude Code Auto Mode防御。
AI Agents Gone Rogue: How OpenAI, Anthropic & Meta Models Accidentally Hacked Real Companies in 2026 — and What Claude Code Auto Mode Does About I…
7
Prompt Injections for Defense
从安全视角拆解提示注入的新用途：借恶意示例反向测试模型防线，揭示 LLM 在危险内容与审查触发词面前的响应机制。
This seems to work : Researchers from Tracebit on Monday said they found that placing prompt injections alongside passwords, cryptographic keys, and o…
8
Towards Realistic Guarantees: A Probabilistic Certificate for SmoothLLM
针对SmoothLLM防御框架提出概率性认证方法，突破传统确定性保证局限，为LLM对抗鲁棒性提供更贴近实际场景的量化评估。
arXiv:2511.18721v4 Announce Type: replace-cross Abstract: The SmoothLLM defense provides a certification guarantee against jailbreaking attacks, but i…
9
SIEVE: Selective Integrity Verification and Escalation for Defending LLM Agents against Indirect Prompt Injection
一套选择性完整性验证机制，精准拦截针对LLM智能体的间接提示注入攻击，安全研究新思路。
arXiv:2512.06716v3 Announce Type: replace Abstract: Large Language Models (LLMs) are increasingly used as the core of agentic systems due to their str…
10
细菌如何控制“自杀”防御？哈工大团队揭示细菌逆转录子双重抑制机制
IT之家 7 月 31 日消息，哈尔滨工业大学生命科学和医学学部黄志伟教授团队最新研究揭示了细菌逆转录子 Ec78 系统的双重抑制机制，阐明了细菌如何通过两道“保险锁”控制毒素蛋白，实现对噬菌体入侵的快速响应。相关成果已于 7 月 31 日发表于《美国科学院院刊》。 逆转录子是细菌体内重要的免疫系统…
11
Adversarial Code Obfuscation for Defending Against LLM-Based Analysis
用对抗性代码混淆技术防御大语言模型分析，Acoda方法为LLM安全带来全新视角
Article URL: https://arxiv.org/abs/2606.11755 Comments URL: https://news.ycombinator.com/item?id=49109869 Points: 2 # Comments: 1
12
A fundamental flaw leaves LLMs strikingly vulnerable to attack
LLM存在根本性缺陷，极易被攻击，一个简单提示就能绕开安全限制。
It is impossible to make large language models fully secure against hacks because of a fundamental flaw in how they work, a team of researchers argue …
13
ToxScreen: Detecting Whether an LLM Has Been Poisoned
针对大模型投毒攻击的检测新方法ToxScreen，通过分析模型行为判断是否被恶意篡改
arXiv:2607.26849v1 Announce Type: cross Abstract: As large language models (LLMs) are deployed in high-stakes domains, adversaries may poison training…
14
美国三州将试点校园无人机防御系统，最快 15 秒响应枪击
IT之家 7 月 29 日消息，美国学校即将加入无人机军备竞赛行列。Mithril Defense 公司推出的“校园守护天使”（Campus Guardian Angel）无人机计划将在今年年底前进入至少 3 个州的 9 所学校进行测试。 据《华盛顿邮报》报道，该试点项目将在佛罗里达州、佐治亚州和科…
15
Build a Basic AI Agent From Scratch: Security III
手把手教你为AI Agent构建三层安全防护，从SSRF到shell策略，代码可克隆实操
Previous parts of Build a Basic AI Agent From Scratch : Basic Agent Tools Long Task Planning Human in the Loop & Security Security II You can find…
16
Disrupting supply chain attacks on npm and GitHub Actions
GitHub安全团队揭秘如何阻断npm与Actions供应链攻击链，守护开源生态从源头做起。
Explore the changes we've shipped across npm and GitHub Actions over the past few months to disrupt supply chain attack techniques and limit their imp…
17
Robust Critics: Defending LLMs Against Multi-Turn Attacks
让大模型不再“被套路”：多轮攻击通过动态推断用户意图实现防御。
arXiv:2607.20472v1 Announce Type: new Abstract: When a user asks a language model something harmful, is it a genuine attack or a misunderstood but wel…
18
Broken Gates: Re-evaluating Web Bot Defenses in the Age of LLM Agents
LLM代理正以惊人效率攻破传统网页机器人防御，这篇论文系统揭示了当前安全机制的致命盲区。
arXiv:2607.18659v1 Announce Type: cross Abstract: LLM-based browser agents are rapidly changing the threat landscape for web security. Unlike traditio…
19
CluCERT: Certifying LLM Robustness via Clustering-Guided Denoising Smoothing
通过聚类引导去噪平滑为LLM鲁棒性提供形式化认证，为AI安全理论注入新思路
arXiv:2512.08967v2 Announce Type: replace Abstract: Recent advancements in Large Language Models (LLMs) have led to their widespread adoption in daily…
20
SociaLLM Engineering: On Manipulating AI Agents and what we can do about it
AI Agent正成为新型社会工程攻击目标，本文深入剖析攻击原理并给出防御建议，安全从业者必读。
Article URL: https://cephalosec.com/blog/sociallm-engineering-old-tricks-ai-agents-are-the-new-victims/ Comments URL: https://news.ycombinator.com/ite…