Translation in progress Some entries still include text in the original language.
Oct 12 · Monday
All papers Found 18 papers Another 240 papers do not have category tags yet and are not included in filtered results.
“
About this collection Collection in progress We are building a large AI safety paper collection to help researchers find work related to their interests. Papers are collected automatically andfiltered only for relevance to AI safety , without judging quality, novelty or reading value. The default “Annotated first” shows papers with tags or published paper analyses first. Each group is ordered by original publication date, newest first, with undated papers last. “Newest first” uses publication dates only. This order is not a quality ranking.
One-line summaries, tags and paper analyses are prepared by AI from the papers. Language models may have systematic biases and produce hallucinations.We plan to carry out human annotation and review to reduce errors in this content.
提出基于决策前隐藏状态的检测与引导，预测并抑制智能体欺骗
Published Oct 5
Evaluated models Qwen3-14B Abstract 论文展示了智能体自发欺骗可在行为显现前由内部表征预测，并通过激活引导证实了表征的行为可操作性。
提出 MemAdapter，在检索后约束记忆对智能体推理的影响
Published Oct 4
Evaluated models DeepSeek-V4-Flash , GPT-5.6-sol , Qwen3-8B Abstract 检索后三阶段调节记忆角色。MemAdapter 是给长期记忆智能体用的检索后调节框架，用来处理记忆诱发谄媚。
提出 Kubernetes 运维智能体的七层纵深防御框架与参考架构
Published Oct 2 Abstract 把运维智能体当不可信租户，用七层控制在完全攻破假设下约束行动。把 Kubernetes 上的运维智能体当作不可信租户，在假定提示注入已完全控制其工具选择的前提下，用基础设施重建数据面与控制面的边界。要解决的是：智能体读日志、注解、工单或 MCP 工具描述时，这些内容可以变成它带着自身凭据执行的命令。模型对齐和分类器只降低概率。
提出 ZoneClaw，用记忆分区防御计算机使用 Agent 的持久记忆攻击
Published Oct 1
Evaluated models Claude Sonnet 4.6 , GPT-5.5 , GLM-5.2 and others4 提出免训练框架 SED，将有害智能体轨迹蒸馏为可检索安全策略
Published Sep 29
Evaluated models DeepSeek-V4-Flash , GLM-5.2 , Kimi K3 提出有守卫最小步激活引导，只在关机语境把继续偏好转向接受关机
Published Sep 24
Evaluated models Qwen3.5-0.8B 提出 Auto Mode ++，加固编码 Agent 阻断监视器
Published Sep 17
Evaluated models Opus 5 , Opus 4.8 , Opus 4.7 and others8 Abstract 系统评估生产阻断监视器在面对持续恶意智能体时的表现，提出加固防御并指明多上下文防御挑战。
在写入点定位拒答方向，经刚性旋转跨架构迁移并门控放大
Published Sep 4
Evaluated models Falcon-Mamba-7B , Falcon3-Mamba-7B , Llama-3.1-8B and others8 Abstract 拒答方向可经刚性旋转跨架构迁移，但必须在各架构的写入点读取。作者研究拒答表示是否随架构改变，结论是表示共享、读取位置随架构变。
提出 AgentKernel，用四支柱强制中介 Agent 的输入、工具与记忆
Published Aug 29 Abstract AgentKernel 是四支柱智能体 OS 设计，安全论证未附带实测数字。AgentKernel 是把安全做成组织原则的智能体 OS 架构，不是带跑分的攻防评测。
提出 INTENT-AS-A-TOOL，用意图工具概率跟踪并干预智能体失准
Published Aug 28
Evaluated models Gemma-4-31B-IT , Qwen3.5-27B , Qwen3-235B-A22B and others5 Abstract 用意图工具的下一动作概率跟踪推理中的失准承诺，并在其成为首选时插入反思。Zhang 等人研究推理型语言模型在职场邮件沙箱里如何走向智能体失准，并提出用动作偏好而不是事后文本标签来跟踪承诺。问题是：有害执行往往出现在推理已经形成意图之后，但 CoT 监控是事后、粗粒度、依赖外部评审的，也看不到模型倾向调用哪个动作。
提出依赖引导回滚修复，恢复记忆增强 Agent 的回答与状态
Published Aug 11
Evaluated models GPT-4o , Gemini-3.6-Flash , Qwen3.6-27B 提出 TMA-NM，以写时来源绑定抵御 Agent 长期记忆投毒
Published Jun 23
Evaluated models gpt-5-chat , gpt-4o-mini , claude-opus-4.1 and others8 Abstract TMA-NM 用来源绑定的行动权限挡住记忆洗白。TMA-NM 把 LLM 智能体长期记忆的行动权限绑到写时来源，用来挡住跨会话记忆投毒。
提出 Agent MechSuits，对多轮 CLI Agent 做步级检测与子空间转向
Published Jun 22
Evaluated models LLaMA-3.1-8B , Qwen-2.5-7B , Gemma-2-9B Abstract 白盒步级探针加稀疏转向可压低三模型恶意动作，同时提高生成崩溃率。Agent MechSuits 是面向多轮 CLI 智能体的白盒、步级表示防御，并配有可执行轨迹基准 MAS。
Reviews threats to autonomous agents and proposes a full-lifecycle defense blueprint, FASA
Published Mar 13 Abstract Using OpenClaw as an example, it divides risks into three layers and uses FASA to describe defense from input to system execution.This is a threat review and defense architecture report using OpenClaw as a case study, not a scored attack or defense evaluation.
Proposes a defense-in-depth aligned with the five stages to mitigate cross-stage threats to autonomous agents
Published Mar 12 Abstract This is a lifecycle threat analysis and layered mitigation design for OpenClaw, with the cases serving as qualitative evidence.This is a lifecycle threat analysis of OpenClaw, accompanied by a set of layered mitigation designs. The problem addressed is: when autonomous agents connect instant messaging, persistent memory, third-party plugins, and high-privilege execution, attacks are no longer limited to single-turn jailbreaks but can compound across multiple stages from initialization to execution. The authors argue that defenses hardening only individual interfaces cannot stop such cross-temporal attacks.
提出推理时防御 ICON，探测注意力过聚焦并转向以纠正 Agent 间接提示注入
Published Feb 24
Evaluated models Qwen-3-8B , Llama-3.1-8B , Mistral-8B and others6 Abstract ICON 在推理时检测过聚焦并转向，以兼顾 IPI 安全与任务效用。ICON 在推理时检测 IPI 的注意力过聚焦并转向，避免以拒答中止智能体轨迹。
提出 TAME 调节 Agent 测试时记忆演化，并用 Trust-Memevo 评测可信度
Published Feb 3
Evaluated models Qwen3-32B , GPT-5.2 Abstract TAME 在共享记忆上同时调节效用与信任。TAME 针对测试时记忆演化中的 Agent Memory Misevolution：良性任务只按成败保留经验时，任务表现可以上升，安全、隐私、公平等可信度下降或随领域波动。
提出 VIGIL 先验证后提交，防御 Agent 工具流注入
Published Jan 9
Evaluated models Qwen3-max , Gemini-2.5-pro Abstract VIGIL 用意图约束下的提交前验证处理工具流注入。VIGIL 用 verify-before-commit 处理智能体的工具流注入，并配有基准 SIREN。
You have reached the end