Translation in progress Some entries still include text in the original language.

Oct 12 · Monday

All papers

Found 11 papers

Filter1

Attack type

Jailbreaks9

Impact

Harmful content8

System type

Models and APIs7 AI Agent3

Attacker access

Black-box11

Attack surface / entry point

User input11 Images, audio and video3

Layer

Prompt layer7 Application layer3

System component

Guardrails and judges6 Image generation3 Multimodal4

Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 265papers have no tags yet and are not included in filtered results.

Attacker access：Black-box Clear all

Another 265papers do not have category tags yet and are not included in filtered results.

Evaluation and benchmarksarXiv 2610.11112Oct 8
EpiReal-Bench：商用图像生成模型虚假声明红队基准
构建 EpiReal-Bench，评测图像生成器将虚假主张渲染为可信证据
arXiv
2610.11112
Published
Oct 8
Layer
Prompt layer
Setting
Models and APIs
Attacker
Black-box
Evaluated models
GPT-Image-2, Nano Banana 2, Grok Imagen 2.0 and others4
- AttacksJailbreaks
- ComponentImage generation
Abstract四款商用图像模型能把假主张画成可信证据。
EpiReal-Bench 用来量商用图像生成器会不会把虚假现实主张画成可信视觉证据。
Paper analysisRead analysis
Evaluation and benchmarksarXiv 2610.03434Oct 2
Persona Guardrail：面向智能体系统的生产级防御框架
提出 Persona Guardrail，强制客服智能体守住功能边界
arXiv
2610.03434
Published
Oct 2
Layer
Application layer
Setting
AI Agent
Attacker
Black-box
Evaluated models
Qwen3-30B-A3B, Qwen3-VL-30B, Qwen3-8B and others5
- DefensesDetection and filtering
- AttacksJailbreaks
- ComponentGuardrails and judges
AbstractPersona Guardrail 用功能边界做同步输入输出校验，并在 PAGE 与生产延迟预算下给出评测。
Persona Guardrail 是客服向智能体的同步运行时防御：先写明功能边界，再决定放行或拦截，而不是只检测某一攻击类。
Paper analysisRead analysis
Evaluation and benchmarksarXiv 2609.33640Sep 27
SafeMol：分子多模态模型的双模态安全对齐
提出 SafeMol，联合对齐分子多模态模型的文本与图输入
arXiv
2609.33640
Published
Sep 27
Layer
Training and data layer
Setting
Models and APIs
Attacker
Black-box
Evaluated models
Omni-Mol v2, ReactXT, SafeMol
- DefensesModel hardening
- AttacksJailbreaks
AbstractSafeMol 用 MMD 和双分类头对齐分子多模态模型，在 Omni-Mol v2 上压低图条件攻击成功。
SafeMol 研究分子多模态模型在危险分子相关请求上的安全对齐，并同时看过拒和任务效用。
Paper analysisRead analysis
Evaluation and benchmarksarXiv 2609.22724Sep 19
MATE：面向移动 Agent 的策略感知安全审计方法
提出移动 Agent 轨迹审计器 MATE，判断轨迹是否违反可编辑安全策略
arXiv
2609.22724
Published
Sep 19
Layer
Application layer
Setting
Browser and computer use
Attacker
Black-box
Evaluated models
MATE-0.5B, MATE-1.5B, MATE-3B
- DefensesMonitoring and auditing
- AttacksMalicious use
- RiskDangerous agent actions
- ComponentMonitors
AbstractMATE 用合成轨迹微调轻量审计器，在可编辑策略下判断移动智能体轨迹是否违规并给出解释。
MATE 是面向移动 GUI 智能体的轻量策略条件轨迹审计器，在给定自然语言安全策略时判断轨迹是否违规并解释原因。
Paper analysisRead analysis
Evaluation and benchmarksarXiv 2609.09798Sep 9
CS-Guard：首个面向代码生成安全的 LLM 护栏评测基准
提出 CS-Guard 基准，评测代码生成护栏对虚构场景攻击的防御
arXiv
2609.09798
Published
Sep 9
Layer
Prompt layer
Setting
coding agent
Attacker
Black-box
Evaluated models
CodeLlama-13B-Instruct, DeepSeekCoder-V2-Lite-16B, GPT-OSS-20B and others7
- AttacksJailbreaks
- ComponentGuardrails and judges
Abstract论文提出了代码安全护栏基准 CS-Guard，评估发现现有护栏对虚构场景攻击和代码级恶意请求防御较弱。
Paper analysisRead analysis
Evaluation and benchmarksarXiv 2609.07216Sep 7
DIVA：利用视觉锚点对视频生成模型实施多模态越狱
提出 DIVA，将有害意图拆成视觉锚点与运动文本以越狱视频生成模型
arXiv
2609.07216
Published
Sep 7
Layer
Prompt layer
Setting
Models and APIs
Attacker
Black-box
Evaluated models
Wan, CogVideoX, LTX-Video and others9
- AttacksJailbreaks
- TechniqueCross-modal payloads
- ComponentImage generation
AbstractDIVA 用视觉锚点抑制语义漂移做视频越狱，并给出多条件安全基准 TI2VSafetyBench。
DIVA 是针对带参考图的视频生成模型的免训练 black-box 越狱：用一致性机制本身堵住语义漂移这条安全逃逸路径。。
Paper analysisRead analysis
Evaluation and benchmarksarXiv 2608.17360Aug 18
Fair-ASR：在共享目标调用预算下重新评测黑盒越狱攻击
在共享目标调用预算下重评黑盒越狱并提出 ReCode
arXiv
2608.17360
Published
Aug 18
Layer
Prompt layer
Setting
Models and APIs
Attacker
Black-box
Evaluated models
Llama-3.1-8B-Instruct, Llama-3.1-70B-Instruct, gpt-oss-20b and others8
- AttacksJailbreaks
Paper analysisRead analysis
Evaluation and benchmarksarXiv 2606.18673Jun 17
理解并缓解真实世界 LLM 应用中的提示词泄露攻击
提出软提示防御 AREA，重锚定注意力以缓解系统提示泄露
arXiv
2606.18673
Published
Jun 17
Layer
Application layer
Setting
LLM applications
Attacker
Black-box
Evaluated models
Llama-2-7B-chat-hf, Llama-3.1-8B-Instruct, Mistral-7B-Instruct-v0.3 and others8
- DefensesInference-time intervention
- AttacksExtraction and theft
Abstract作者测量商用提示词泄漏，提出 attention drift，并用 AREA 在可用性与抗泄漏间取折中。
作者研究真实 LLM 应用中的 system prompt 泄漏，并给出不改权重的缓解方法 AREA。
Paper analysisRead analysis
Evaluation and benchmarksarXiv 2604.01444Apr 2
烹饪风险：大语言模型食品安全风险的基准测试与降低
提出 FoodGuardBench 与护栏 FoodGuard-4B，评测并降低食品危害越狱
arXiv
2604.01444
Published
Apr 2
Layer
Prompt layer
Setting
Models and APIs
Attacker
Black-box
Evaluated models
Claude-3.7-Sonnet, GPT-4o, GPT-4.1 and others10
- DefensesDetection and filtering
- AttacksJailbreaks
- TechniqueAutomated search
- ComponentGuardrails and judges
Paper analysisRead analysis
Evaluation and benchmarksarXiv 2602.10179Feb 11
研究者提出视觉中心越狱攻击 VJA
提出纯视觉越狱攻击 VJA，把有害编辑指令编码进图像
arXiv
2602.10179
Published
Feb 11
Layer
Prompt layer
Setting
Models and APIs
Attacker
Black-box
Evaluated models
Nano Banana Pro, GPT Image 1.5, Qwen-Image-Edit and others9
- AttacksJailbreaks
- TechniqueCross-modal payloads
- ComponentImage generation
Paper analysisRead analysis
Evaluation and benchmarksarXiv 2603.20198Feb 5
研究者提出 MM-Plan 多模态越狱框架
提出 MM-Plan 多模态越狱框架，自动规划多轮图像编辑以越狱视觉独占目标
arXiv
2603.20198
Published
Feb 5
Layer
Prompt layer
Setting
Models and APIs
Attacker
Black-box
Evaluated models
Llama-3.2-11B, InternVL3-8B, Qwen3-VL-8B and others8
- AttacksJailbreaks
- TechniqueAutomated search
- ComponentVision
Abstract作者称全局规划能攻破对单轮攻击更稳的前沿 MLLM。
MM-Plan 是面向 Visual Exclusivity 的 black-box 多模态红队方法，并配有人工基准 VE-Safety。
Paper analysisRead analysis

You have reached the end

AI Safety Daily

Read the latest issueOne issue per day, updated as events develop