auto: new LLM security papers (run #16)

[github-actions]

Automated Paper Fetch

New papers discovered from arXiv, Semantic Scholar, and CrossRef.

All entries are marked reviewed: false -- please review before merging.

New Entries

--- New Entries ---
- [llmsec-2026-00206] When Agents Remember Too Much: Memory Poisoning Attacks on Large Language Model Agents
  https://arxiv.org/abs/2607.06595
  Categories: memory-security
- [llmsec-2026-00207] Antaeus: Hunting Repository-Level Logic Vulnerabilities via Context-Grounded LLM Reasoning
  https://arxiv.org/abs/2607.01138
  Categories: agentic-threats
- [llmsec-2026-00208] A Lifecycle and Application-Stack Survey of Large Language Model Vulnerabilities: Attacks, Risks, Defenses, and Open Problems
  https://arxiv.org/abs/2606.31639
  Categories: autonomous-operations, survey
- [llmsec-2026-00209] Toward Secure and Reliable PDDL Formalization of Large Language Models with Planner-in-the-Loop Feedback
  https://arxiv.org/abs/2606.29700
  Categories: benchmarks
- [llmsec-2026-00210] An Empirical Evaluation of Prompt Injection Vulnerabilities in Large Language Models Across Multilingual and Obfuscated Attack Scenarios
  https://arxiv.org/abs/2606.29602
  Categories: prompt-injection
- [llmsec-2026-00211] Memory as an Attack Surface in LLM Agents: A Study on Multiple-Choice Question Answering
  https://arxiv.org/abs/2606.29030
  Categories: tool-use-security
- [llmsec-2026-00212] FlipGuard: Defending Large Language Models Against Quantization-Conditioned Backdoor Attacks
  https://arxiv.org/abs/2606.28962
  Categories: data-poisoning
- [llmsec-2026-00213] RIPA: Sensory-Vector Prompt Injection Attacks on LLM-Controlled ROS 2 Robots
  https://arxiv.org/abs/2606.28649
  Categories: prompt-injection
- [llmsec-2026-00214] Robust Harmful Features Under Jailbreak Attacks: Mechanistic Evidence from Attention Head Specialization in Large Language Models
  https://arxiv.org/abs/2606.28153
  Categories: jailbreaking, guardrails
- [llmsec-2026-00215] LLM agents security duality: a comprehensive survey of self-security and empowered cybersecurity
  https://arxiv.org/abs/2606.28450
  Categories: survey
- [llmsec-2026-00216] Representation Matters: An Empirical Study of Program Representations for LLM Vulnerability Reasoning
  https://arxiv.org/abs/2606.25356
  Categories: benchmarks
- [llmsec-2026-00217] HelpBench: Assessing the Ability of LLMs to Provide Privacy, Safety, and Security Advice
  https://arxiv.org/abs/2606.24819
  Categories: benchmarks
- [llmsec-2026-00218] Securing LLM-Agent Long-Term Memory Against Poisoning: Non-Malleable, Origin-Bound Authority with Machine-Checked Guarantees
  https://arxiv.org/abs/2606.24322
  Categories: memory-security
- [llmsec-2026-00219] Evaluating LLMs for Real-World Web Vulnerability Detection
  https://arxiv.org/abs/2606.21397
  Categories: benchmarks
- [llmsec-2026-00220] Local LLM Agents as Vulnerable Runtimes:A Source-Code Audit of the Agent Runtime Layer
  https://arxiv.org/abs/2606.21071
  Categories: prompt-injection
- [llmsec-2026-00221] Honeyquest for LLMs: Rethinking Cyber Deception for AI Attackers
  https://arxiv.org/abs/2606.21037
  Categories: benchmarks
- [llmsec-2026-00222] Calibration Without Comprehension: Diagnosing the Limits of Fine-Tuning LLMs for Vulnerability Detection in Systems Software
  https://arxiv.org/abs/2606.20502
  Categories: benchmarks
- [llmsec-2026-00223] The Consistency Dilemma in LLMs: Generator-Evaluator Agreement and Vulnerability to Mistakes
  https://arxiv.org/abs/2606.30653
  Categories: agentic-threats
- [llmsec-2026-00224] How Much Can We Trust LLM Search Agents? Measuring Endorsement Vulnerability to Web Content Manipulation
  https://arxiv.org/abs/2606.16821
  Categories: benchmarks, threat-modeling
- [llmsec-2026-00225] SkillVetBench: LLM-as-Judge for Multi-Dimensional Security Risk Evaluation in Open-Source LLM Agent Skills
  https://arxiv.org/abs/2606.15899
  Categories: agent-architecture
- [llmsec-2026-00226] Let Them Steal: Trapping Large Language Model Extraction Attacks with Knowledge Honeypot
  https://arxiv.org/abs/2606.15810
  Categories: model-extraction
- [llmsec-2026-00227] AttackonCTF: Defending Hardware Security Competition Benchmarks in the Age of LLMs
  https://arxiv.org/abs/2606.15809
  Categories: benchmarks
- [llmsec-2026-00228] AutoDojo: Adaptive Black-Box Attacks Reveal the Limits of IPI Defenses and Task-Specification Effects in LLM Agents
  https://arxiv.org/abs/2606.15057
  Categories: prompt-injection
- [llmsec-2026-00229] From Shield to Target: Denial-of-Service Attacks on LLM-Based Agent Guardrails
  https://arxiv.org/abs/2606.14517
  Categories: prompt-injection, jailbreaking, guardrails, autonomous-operations
- [llmsec-2026-00230] SkillMutator: Benchmarking and Defending Language-and-Code Cross-modal Attacks on LLM Agent Skills
  https://arxiv.org/abs/2606.14154
  Categories: benchmarks
- [llmsec-2026-00231] Can Open-Source LLM Agents Replace Static Application Security Testing Tools? An Empirical Assessment
  https://arxiv.org/abs/2606.11672
  Categories: agentic-threats
- [llmsec-2026-00232] MPC-Patch-Bench: Security-Aware LLM Code Patch for Multi-Party Computation
  https://arxiv.org/abs/2606.11416
  Categories: cryptographic-controls, benchmarks
- [llmsec-2026-00233] Toward Secure LLM Agents: Threat Surfaces, Attacks, Defenses, and Evaluation
  https://arxiv.org/abs/2606.10749
  Categories: agentic-threats
- [llmsec-2026-00234] Multi-Agent Firewall Architecture for Privacy Protection of Sensitive Data in Interactions with Language Models
  https://arxiv.org/abs/2607.08282
  Categories: agent-architecture
- [llmsec-2026-00235] Prismata: Confining Cross-Site Prompt Injection in Web Agents
  https://arxiv.org/abs/2607.08147
  Categories: prompt-injection
- [llmsec-2026-00236] Efficient Safety Alignment of Language Models via Latent Personality Traits
  https://arxiv.org/abs/2607.07918
  Categories: adversarial-examples, guardrails
- [llmsec-2026-00237] Mechanistic Interpretability of LLM Jailbreaks via Internal Attribution Graphs
  https://arxiv.org/abs/2607.07903
  Categories: jailbreaking, adversarial-examples
- [llmsec-2026-00238] Open Models, Open Risks: Measuring Unsafe Generation in Text-to-Image Models In the Wild
  https://arxiv.org/abs/2607.07827
  Categories: jailbreaking
- [llmsec-2026-00239] Beware of Agentic Botnets: Scalable Untargeted Promptware Attacks via Universal and Transferable Adversarial HalluSquatting
  https://arxiv.org/abs/2607.07433
  Categories: prompt-injection, agentic-threats, threat-modeling
- [llmsec-2026-00240] Untrusted Content Masking for Web Agents with Security Guarantees
  https://arxiv.org/abs/2607.05277
  Categories: prompt-injection
- [llmsec-2026-00241] Agent Data Injection Attacks are Realistic Threats to AI Agents
  https://arxiv.org/abs/2607.05120
  Categories: prompt-injection, agentic-threats
- [llmsec-2026-00242] DualView: Preventing Indirect Prompt Injection in Personal AI Agents
  https://arxiv.org/abs/2607.03821
  Categories: prompt-injection
- [llmsec-2026-00243] Overloading Large Vision-Language Models for Jailbreaking
  https://arxiv.org/abs/2607.02961
  Categories: jailbreaking
- [llmsec-2026-00244] HARC: Coupling Harmfulness and Refusal Directions for Robust Safety Alignment
  https://arxiv.org/abs/2607.00572
  Categories: jailbreaking, guardrails
- [llmsec-2026-00245] Beyond the Prompt: Jailbreaking Function-Calling LLMs via Simulated Moderation Traces
  https://arxiv.org/abs/2607.00481
  Categories: jailbreaking
- [llmsec-2026-00246] Securing the AI Agent: A Unified Framework for Multi-Layer Agent Red Teaming
  https://arxiv.org/abs/2606.31227
  Categories: red-teaming
- [llmsec-2026-00247] Security--Fidelity Tradeoffs: The Hidden Cost of Prompt Injection Defense
  https://arxiv.org/abs/2606.30783
  Categories: prompt-injection, benchmarks
- [llmsec-2026-00248] Understanding and Evaluating Claw-like Agent Security Through a Computer-Systems Lens
  https://arxiv.org/abs/2606.30755
  Categories: agentic-threats, benchmarks
- [llmsec-2026-00249] Forensic Trajectory Signatures for Agent Memory Poisoning Detection
  https://arxiv.org/abs/2606.30566
  Categories: memory-security
- [llmsec-2026-00250] On the Inseparability of Instructions and Data in Shared-Embedding Sequence Models
  https://arxiv.org/abs/2606.27567
  Categories: prompt-injection, access-control
- [llmsec-2026-00251] Jailbreaking for the Average Jane: Choosing Optimal Jailbreaks via Bandit Algorithms for Automatically Enhanced Queries
  https://arxiv.org/abs/2606.26936
  Categories: jailbreaking
- [llmsec-2026-00252] MIRROR: Novelty-Constrained Memory-Guided MCTS Red-Teaming for Agentic RAG
  https://arxiv.org/abs/2606.26793
  Categories: prompt-injection, agentic-threats, red-teaming, benchmarks
- [llmsec-2026-00253] Agents That Know Too Much: A Data-Centric Survey of Privacy in LLM Agents
  https://arxiv.org/abs/2606.26627
  Categories: survey
- [llmsec-2026-00254] Adversarial Diffusion Across Modalities: A Fusion Survey of Attacks, Defenses, and Evaluation for Text, Vision, and Vision-Language Models
  https://arxiv.org/abs/2606.26566
  Categories: jailbreaking, benchmarks, survey, threat-modeling
- [llmsec-2026-00255] Adaptive Evaluation of Out-of-Band Defenses Against Prompt Injection in LLM Agents
  https://arxiv.org/abs/2606.26479
  Categories: prompt-injection, benchmarks
- [llmsec-2026-00256] RAS: Measuring LLM Safety Through Refusal Alignment
  https://arxiv.org/abs/2606.25750
  Categories: jailbreaking, guardrails
- [llmsec-2026-00257] How Reliable Is Your Jailbreak Judge? Calibration and Adversarial Robustness of Automated ASR Scoring
  https://arxiv.org/abs/2606.25487
  Categories: prompt-injection, jailbreaking
- [llmsec-2026-00258] PixJail: Self-Evolving Paper-to-Pipeline Reproduction for Text-to-Image Jailbreak Evaluation
  https://arxiv.org/abs/2606.24081
  Categories: jailbreaking, benchmarks
- [llmsec-2026-00259] TROPT: An Open Framework for Unifying and Advancing Discrete Text Optimization
  https://arxiv.org/abs/2606.23496
  Categories: jailbreaking, red-teaming
- [llmsec-2026-00260] Detecting Malicious Agent Skills in the Wild using Attention
  https://arxiv.org/abs/2606.23416
  Categories: agentic-threats
- [llmsec-2026-00261] DE-FIVE: Detecting Malicious Image Prompts via Fourier Features and Image Vector Embeddings
  https://arxiv.org/abs/2606.22779
  Categories: prompt-injection, adversarial-examples
- [llmsec-2026-00262] The Geometry of Refusal: Linear Instability in Safety-Aligned LLMs
  https://arxiv.org/abs/2606.22686
  Categories: guardrails
- [llmsec-2026-00263] Confidently Wrong: Severity-Aware Calibration of Prompt-Injection Detectors under Attack Shift
  https://arxiv.org/abs/2606.22659
  Categories: benchmarks
- [llmsec-2026-00264] Safe to Check, Unsafe to Use: Relinking at the Compression Boundary of LLM Agents
  https://arxiv.org/abs/2606.21732
  Categories: prompt-injection
- [llmsec-2026-00265] Toward Open Weight Models Without Risks: Separating Public and Private Capabilities in LLMs
  https://arxiv.org/abs/2606.21638
  Categories: jailbreaking
- [llmsec-2026-00266] AgenticOS: An Intent-Oriented Secure Operating System Architecture for Autonomous AI Agents
  https://arxiv.org/abs/2606.21129
  Categories: prompt-injection, agentic-threats, access-control, tool-use-security, autonomous-operations
- [llmsec-2026-00267] Scalable Hierarchical Attention Transformers for Multi-Turn Jailbreak Detection in Long Conversations
  https://arxiv.org/abs/2606.21082
  Categories: jailbreaking
- [llmsec-2026-00268] Think Twice Before You Act: Protecting LLM Agents Against Tool Description Poisoning via Isolated Planning
  https://arxiv.org/abs/2606.20922
  Categories: prompt-injection
- [llmsec-2026-00269] Analyzing Defensive Misdirection Against Model-Guided Automated Attacks on Agentic AI Systems
  https://arxiv.org/abs/2606.20470
  Categories: jailbreaking, agentic-threats
- [llmsec-2026-00270] A Layered Security Framework Against Prompt Injection in RAG-Based Chatbots
  https://arxiv.org/abs/2606.19660
  Categories: prompt-injection, input-filtering
- [llmsec-2026-00271] Amplify, Don't Create: Temporal Accumulation for Slow-Burn Prompt Injection
  https://arxiv.org/abs/2606.20746
  Categories: prompt-injection
- [llmsec-2026-00272] CodeSentinel: A Three-Layer Defense Against Indirect Prompt Injection in Code Contexts
  https://arxiv.org/abs/2606.19235
  Categories: prompt-injection
- [llmsec-2026-00273] The Gate Is Only as Honest as Its Contracts: ContractGuard for the Contract Layer of Risk-Aware Causal Gating
  https://arxiv.org/abs/2606.18550
  Categories: prompt-injection, access-control, tool-use-security
- [llmsec-2026-00274] SafeClawBench: Separating Semantic, Audit-Evidence, and Sandbox Harm in Tool-Using LLM Agents
  https://arxiv.org/abs/2606.18356
  Categories: agentic-threats, sandboxing-isolation, benchmarks
- [llmsec-2026-00275] A Red-Team Study of Anthropic Fable 5 & Opus 4.8 Models
  https://arxiv.org/abs/2606.18193
  Categories: jailbreaking, red-teaming
- [llmsec-2026-00276] Adversarial Pragmatics for AI Safety Evaluation: A Benchmark for Instruction Conflict, Embedded Commands, and Policy Ambiguity
  https://arxiv.org/abs/2607.01153
  Categories: agentic-threats, benchmarks
- [llmsec-2026-00277] Yuvion LLM: An Adversarially-Aware Large Language Model for Content And AI Safety
  https://arxiv.org/abs/2606.27632
  Categories: tool-use-security
- [llmsec-2026-00278] MultiTurnPSB: Evaluating Multi-Turn Jailbreak Attacks an dClassifier-Based Defenses for Medical AI Safety
  https://arxiv.org/abs/2606.02630
  Categories: jailbreaking, adversarial-examples
- [llmsec-2026-00279] AI Security Research Should Better Incentivize Defense Research
  https://arxiv.org/abs/2605.23448
  Categories: membership-inference, federated-learning
- [llmsec-2026-00280] The Poisoned Chalice of LLM Evaluation Report
  https://arxiv.org/abs/2607.07481
  Categories: benchmarks
- [llmsec-2026-00281] ReShift: Aha-Moment-Driven Reasoning-Level Backdoor Attacks on Vision-Language Models
  https://arxiv.org/abs/2607.00361
  Categories: data-poisoning
- [llmsec-2026-00282] Curvature-Guided Module Localization for Low-Rank Detoxification of Backdoored Large Language Models
  https://arxiv.org/abs/2606.30899
  Categories: data-poisoning
- [llmsec-2026-00283] Theory of Continual Learning Against Data Poisoning Attacks
  https://arxiv.org/abs/2606.29841
  Categories: data-poisoning
- [llmsec-2026-00284] Why Trust Your Agent? Empirical Security Gains from TRiSM-Guided Agentic Workflows in Healthcare
  https://arxiv.org/abs/2606.28666
  Categories: agentic-threats
- [llmsec-2026-00285] When the Aggregator Cheats: Data-Free Backdoors in Federated LLM-based QA Systems
  https://arxiv.org/abs/2606.27511
  Categories: federated-learning
- [llmsec-2026-00286] Detect, Unlearn, Restore: Defending Text Summarization Models Against Data Poisoning
  https://arxiv.org/abs/2606.26036
  Categories: data-poisoning
- [llmsec-2026-00287] Forget to Improve: On-Device LLM-Agent Continual Learning via Budget-Curated Memory
  https://arxiv.org/abs/2606.25115
  Categories: agentic-threats
- [llmsec-2026-00288] LLM-Based Scientific Peer Review: Methods, Benchmarks, and Reliability Challenges
  https://arxiv.org/abs/2606.25057
  Categories: benchmarks, survey
- [llmsec-2026-00289] The Containment Gap: How Deployed Agentic AI Frameworks Fail Public-Facing Safety Requirements
  https://arxiv.org/abs/2606.12797
  Categories: agentic-threats
- [llmsec-2026-00290] Securing Code Understanding: Detecting Natural Backdoor Vulnerability in Code Language Models
  https://arxiv.org/abs/2606.10846
  Categories: data-poisoning
- [llmsec-2026-00291] MemVenom: Triggered Poisoning of Multimodal Memories in Web Agents
  https://arxiv.org/abs/2606.10742
  Categories: memory-security
- [llmsec-2026-00292] Game-Theoretic Multi-Agent Control for Robust Contextual Reasoning in LLMs
  https://arxiv.org/abs/2606.10322
  Categories: agent-architecture
- [llmsec-2026-00293] Defending Against Malicious Finetuning by Scaling Train-time Adversarial Attacks
  https://arxiv.org/abs/2606.07970
  Categories: adversarial-examples, guardrails
- [llmsec-2026-00294] Steering Vectors are an Adversarial Attack Surface
  https://arxiv.org/abs/2606.05958
  Categories: jailbreaking, data-poisoning, adversarial-examples
- [llmsec-2026-00295] Membrane: A Self-Evolving Contrastive Safety Memory for LLM Agent Defense
  https://arxiv.org/abs/2606.05743
  Categories: jailbreaking, guardrails
- [llmsec-2026-00296] Patcher: Post-Hoc Patching of Backdoored Large Language Models
  https://arxiv.org/abs/2606.02995
  Categories: jailbreaking, data-poisoning, guardrails
- [llmsec-2026-00297] BYORn: Bootstrap Your Own Responses to Defend Large Vision-Language Models Against Backdoor Attacks
  https://arxiv.org/abs/2606.02947
  Categories: data-poisoning
- [llmsec-2026-00298] Hijacking Agent Memory: Stealthy Trojan Attacks Through Conversational Interaction
  https://arxiv.org/abs/2605.29960
  Categories: memory-security
- [llmsec-2026-00299] When Medical Safety Alignment Fails: A Benchmark for Evaluating LLMs on High-Risk Medical Queries
  https://arxiv.org/abs/2606.28332
  Categories: guardrails, benchmarks
- [llmsec-2026-00300] Cordyceps: Covert Control Attacks on LLMs via Data Poisoning
  https://arxiv.org/abs/2605.26595
  Categories: data-poisoning, monitoring-detection
- [llmsec-2026-00301] GradSentry: Gradient Spectral Entropy for Backdoor Sample Filtering in Large Language Model Fine-Tuning
  https://arxiv.org/abs/2605.26574
  Categories: data-poisoning
- [llmsec-2026-00302] Security of OpenClaw Agents: Fundamentals, Attacks, and Countermeasures
  https://arxiv.org/abs/2605.25435
  Categories: autonomous-operations
- [llmsec-2026-00303] Security in the Fine-Tuning Lifecycle of Large Language Models: Threats, Defenses,Evaluation, and Future Directions
  https://arxiv.org/abs/2605.25073
  Categories: data-poisoning, fine-tuning-security, survey
- [llmsec-2026-00304] When the Manual Lies: A Realistic Benchmark to Evaluate MCP Poisoning Attacks for LLM Agents
  https://arxiv.org/abs/2605.24069
  Categories: benchmarks
- [llmsec-2026-00305] Backdooring Masked Diffusion Language Models
  https://arxiv.org/abs/2605.19262
  Categories: data-poisoning
- [llmsec-2026-00306] Be Kind, Rewrite: Benign Projections via Rewriting Defend Against LLM Data Poisoning Attacks
  https://arxiv.org/abs/2605.19147
  Categories: data-poisoning
- [llmsec-2026-00307] Surviving the Unseen: Predictive Defense for Novel Multi-Turn Multimodal Attacks
  https://arxiv.org/abs/2605.18988
  Categories: agentic-threats, guardrails, autonomous-operations
- [llmsec-2026-00308] Not What You Asked For: Typographic Attacks in Household Robot Manipulation
  https://arxiv.org/abs/2605.18593
  Categories: benchmarks
- [llmsec-2026-00309] OEP: Poisoning Self-Evolving LLM Agents via Locally Correct but Non-Transferable Experiences
  https://arxiv.org/abs/2605.18930
  Categories: agentic-threats
- [llmsec-2026-00310] Hidden in Memory: Sleeper Memory Poisoning in LLM Agents
  https://arxiv.org/abs/2605.15338
  Categories: memory-security
- [llmsec-2026-00311] SolarChain-Eval: A Physics-Constrained Benchmark for Trustworthy Economic Agents in Decentralized Energy Markets
  https://arxiv.org/abs/2607.08681
  Categories: agentic-threats, benchmarks, autonomous-operations
- [llmsec-2026-00312] Persuasion Attacks Can Decrease Effectiveness of CoT Monitoring
  https://arxiv.org/abs/2607.08066
  Categories: jailbreaking, monitoring-detection
- [llmsec-2026-00313] Institutional Red-Teaming: Deployment Rules, Not Just Models, Causally Shape Multi-Agent AI Safety
  https://arxiv.org/abs/2607.07695
  Categories: red-teaming, benchmarks, agent-architecture
- [llmsec-2026-00314] Multi-Agent AI Control: Distributed Attacks Hamper Per-Instance Monitors
  https://arxiv.org/abs/2607.07368
  Categories: agent-architecture
- [llmsec-2026-00315] Security and Privacy in Agentic AI: Grand Challenges and Future Directions
  https://arxiv.org/abs/2607.06608
  Categories: agentic-threats
- [llmsec-2026-00316] The Balkanization of Execution-Security Research for AI Coding Agents: Isolation, Access Control, and Time-of-Check-to-Time-of-Use Vulnerabilities
  https://arxiv.org/abs/2607.05743
  Categories: access-control, sandboxing-isolation, human-in-the-loop
- [llmsec-2026-00317] aiAuthZ: Off-Host, Identity-Bound Authorization for AI Agents
  https://arxiv.org/abs/2607.05518
  Categories: access-control
- [llmsec-2026-00318] CAGE-1: Control, Assurance, and Governance Evaluation for Enterprise Agentic AI
  https://arxiv.org/abs/2607.03510
  Categories: agentic-threats
- [llmsec-2026-00319] Securing Multi-Tool AI Agent Chains With Dynamic, Real-Time Compositional Policies
  https://arxiv.org/abs/2607.03423
  Categories: guardrails
- [llmsec-2026-00320] Agentic-SecPBFT: Agentic AI-Driven Proactive Security Framework for Wireless PBFT Consensus in Mobile Ad-Hoc Networks
  https://arxiv.org/abs/2607.03269
  Categories: agentic-threats
- [llmsec-2026-00321] Hardware-Enforced Semantic Coordination for Safety-Critical Real-Time Autonomous Systems
  https://arxiv.org/abs/2607.02376
  Categories: agentic-threats
- [llmsec-2026-00322] Copewell: A Multi-Agent Swarm Architecture for Equitable Mental Wellness Support
  https://arxiv.org/abs/2607.02245
  Categories: agent-architecture
- [llmsec-2026-00323] Janus: a Playground for User-Involved Agentic Permission Management
  https://arxiv.org/abs/2607.01510
  Categories: agentic-threats
- [llmsec-2026-00324] An Agentic AI Framework to Accelerate Scientific Discovery in Plant Phenotyping
  https://arxiv.org/abs/2606.31831
  Categories: agentic-threats
- [llmsec-2026-00325] Verification-Gated Agentic Mission-State Governance for Intelligent Industrial Multi-Robot Systems
  https://arxiv.org/abs/2606.31339
  Categories: agentic-threats
- [llmsec-2026-00326] Tool Use Enables Undetectable Steganography in Multi-Agent LLM Systems
  https://arxiv.org/abs/2606.28425
  Categories: agentic-threats, agent-architecture, tool-use-security, autonomous-operations
- [llmsec-2026-00327] The Unfireable Safety Kernel: Execution-Time AI Alignment for AI Agents and Other Escapable AI Systems
  https://arxiv.org/abs/2606.26057
  Categories: output-moderation, guardrails, access-control
- [llmsec-2026-00328] SkyChain Intelligence: A Blockchain-Secured Multi-Agent DRL Framework for Low-Altitude Embodied Artificial Intelligence
  https://arxiv.org/abs/2606.24193
  Categories: agent-architecture
- [llmsec-2026-00329] Critique of Agent Model
  https://arxiv.org/abs/2606.23991
  Categories: agentic-threats
- [llmsec-2026-00330] RIFT-Bench: Dynamic Red-teaming For Agentic AI Systems
  https://arxiv.org/abs/2606.23927
  Categories: agentic-threats, red-teaming
- [llmsec-2026-00331] Rising From the Ashes: How Agentic AI is Unblocking Challenges in Cybersecurity
  https://arxiv.org/abs/2606.23138
  Categories: agentic-threats
- [llmsec-2026-00332] AgentRiskBOM: A Risk-Scoping Security Bill of Materials for Agentic AI Systems
  https://arxiv.org/abs/2606.21877
  Categories: agentic-threats, human-in-the-loop
- [llmsec-2026-00333] Efficient and Sound Probabilistic Verification for AI Agents
  https://arxiv.org/abs/2606.20510
  Categories: monitoring-detection
- [llmsec-2026-00334] PowerAgentBench-Dyn: A Benchmark for Agentic AI in Power System Dynamic Studies
  https://arxiv.org/abs/2606.20401
  Categories: agentic-threats, benchmarks
- [llmsec-2026-00335] Deontic Policies for Runtime Governance of Agentic AI Systems
  https://arxiv.org/abs/2606.19464
  Categories: agentic-threats, access-control, autonomous-operations
- [llmsec-2026-00336] TxBench-PP: Analyzing AI Agent Performance on Small-Molecule Preclinical Pharmacology
  https://arxiv.org/abs/2606.19245
  Categories: benchmarks
- [llmsec-2026-00337] Code-Augur: Agentic Vulnerability Detection via Specification Inference
  https://arxiv.org/abs/2606.18619
  Categories: agentic-threats
- [llmsec-2026-00338] PromptStudio: A Governance-Aware Agentic AI Framework for Automated Prompt Generation, Optimization, Evaluation, and Lifecycle Management
  https://doi.org/10.22214/ijraset.2026.84092
  Categories: agentic-threats
- [llmsec-2026-00339] Auto-Devops GPT: An Agentic AI Framework for Self-Healing CI/CD Pipelines Using LLM-Based Root Cause Analysis and Reinforcement Learning
  https://doi.org/10.22214/ijraset.2026.84056
  Categories: agentic-threats
- [llmsec-2026-00340] Exploiting large language models in peer review: indirect prompt injection attacks and integrity probes
  https://doi.org/10.1007/s11192-026-05695-x
  Categories: prompt-injection
- [llmsec-2026-00341] Privacy-Preserving GAN for Synthetic Data against Membership Inference Attack
  https://doi.org/10.1145/3820889
  Categories: membership-inference
- [llmsec-2026-00342] CIVIL AND CRIMINAL LIABILITY OF DEEPFAKE AI PLATFORM PROVIDERS IN VIOLATIONS OF THE RIGHT TO REPUTATION
  https://doi.org/10.71131/0rqvry57
  Categories: social-engineering

---

Generated by the weekly paper fetch workflow (run #16, trigger: schedule).