1
Grok exfiltrates user data when malicious instructions are encrypted
加密指令也能攻破AI防线？Grok被曝通过提示注入窃取用户数据，大模型安全再亮红灯。
Cryptographic Context Injection is only the latest way to break an LLM safety guardrail.
2
One Word at a Time: Incremental Completion Decomposition Breaks LLM Safety
逐词分解生成过程，新方法轻松突破大模型安全防线，揭示LLM防护盲区。
arXiv:2604.25921v2 Announce Type: replace Abstract: Large Language Models (LLMs) are trained to refuse harmful requests, yet they remain vulnerable to…
3
A Chinese LLM attacked our lab, so we made it work for us
一场来自DeepSeek的300,000次攻击，反而被作者改造成防御利器，大模型安全攻防的硬核实战。
Article URL: https://jesta.ai/blog/darkreasoning Comments URL: https://news.ycombinator.com/item?id=49158479 Points: 16 # Comments: 6
4
Hackers can use 9 of the most popular AI tools to assemble massive botnets
黑客利用9大AI工具"幻觉"漏洞组建僵尸网络，揭示新型HalluSquatting攻击风险。
"HalluSquatting" weaponizes LLMs' inability to say "I don't know."
5
FloatDoor: Platform-Triggered Backdoors in LLMs
揭示一种由平台机制触发的大模型后门攻击，安全研究者值得关注的新方向。
arXiv:2606.19535v1 Announce Type: cross Abstract: Large language models (LLMs) are increasingly deployed in sensitive settings such as software engine…
6
Grammar-Constrained Decoding Can Jailbreak LLMs into Generating Malicious Code
研究发现利用语法约束解码可诱导LLM生成恶意代码，揭示新型安全漏洞。
arXiv:2606.11817v1 Announce Type: cross Abstract: Large Language Models (LLMs) are increasingly used for code generation, raising concerns that they m…
7
Evolve the Method, Not the Prompts: Evolutionary Synthesis of Jailbreak Attacks on LLMs
不再是优化提示词，而是进化攻击方法本身：一篇提出用进化算法自动合成LLM越狱攻击的论文，攻防研究者必读。
arXiv:2511.12710v2 Announce Type: replace Abstract: Automated red teaming frameworks for Large Language Models (LLMs) have become increasingly sophist…