AI Agent Monitoring

Follow this topic Ask about this topic

AI agent monitoring studies how to observe and assess agents while they use tools, pursue long-running tasks, and interact with other agents, so problems can be detected before they cause harm or waste. Common approaches include trajectory classifiers and LLM judges, early-warning models over trace prefixes, formal runtime checks and action gates, and telemetry or provenance systems that record what happened and why. Some work also uses internal activations, human-facing dashboards, and recovery or intervention mechanisms to complement external observation. A central challenge is making monitors timely and reliable across long contexts and adaptive attacks without excessive false alarms, cost, or loss of useful autonomy.

Latest papers

285 papers51 papers in the last four weeks, up 55% on the four weeks before. 0.3% of all new papers.

Pre-Crime for Agents, at the Last Second: Probing When a Recruited Agent Gives In
Oct 29, 2026Leonidas RaghavLLM Agent Security AI Agent Monitoring
Independent
Soei: Structured Observability of Evidence and Intent for Multi-Agent AI Software Engineering
Oct 29, 2026Mohammed Nayeem Khan, K S Sarun KumarAI Agent Monitoring AI Coding Agents
Intel Corporation
Who Else Is In? Multi-Agent Collusion Initation Via Strategic Vagueness
Oct 27, 2026Pramod Kaushik, Sarath Sivaprasad, Laura Ferrarotti +4Multi-Agent System Security Multi-Agent LLM Systems
University of Trento · Fondazione Bruno Kes sler · CISPA +4
Detecting Covert Collusion between AI Agents through Partner-Outsider Positive Listening
Oct 27, 2026Kevin Yandoka DenamganaïAI Agent Monitoring AI Agent Auditing
Monitor Accuracy Isn't the Answer: A Decision-Value Framework for Evaluating AI Agent Restart Monitors
Oct 27, 2026Siddharth Jain, Venkat N. VedamAI Agent Monitoring AI Agent Evaluation
Intuit
Babysitting AI Agents: A User-Centered Taxonomy and Study of Agentic Delegation Failures
Oct 26, 2026Shravika Mittal, Sunnie S. Y. Kim, Scott Counts +3Agent Failure Analysis AI Agent Reliability
Georgia Institute of Technology, Atlanta, Georgia, USA · Microsoft Research, New York, New York, USA · Microsoft, Redmond, Washington, USA +1
RIDAR: Safeguarding Research Integrity in Autonomous Research Agents via Automated Detection and Runtime Intervention
Oct 22, 2026Jiening Siow, Amy Xin, Junjie Wang +7AI Agent Monitoring Runtime Enforcement for AI Agents
Department of Computer Science and Technology, Tsinghua University · Zhipu AI · School of Information, Renmin University of China
Post-Audit Mirage: Identifiability Limits of Offline Agent Verification
Oct 22, 2026Aman Behera, Payal PayalLLM Agent Verification AI Agent Monitoring
IIT Roorkee
Model Organisms of Collusion: Corrective Transfer in Untrusted Monitors
Oct 22, 2026Juan-Pablo Rivera, Ari Spiesberger, Juan J Vazquez +6AI Control LLM Safety Monitoring
Meridian Cambridge · Arb Research · Redwood Research +3
A Three-Pillar Governance Framework for Safe and Trustworthy AI Agents
Oct 21, 2026Edward C. Cheng, Jeshua Cheng, Alice SiuAI Agent Governance AI Agent Safety
Management Science & Engineering, Stanford University, Stanford, California, United States · AI Laboratory, Inquiryon, Inc., South San Francisco, California, United States · Deliberative Democracy Lab, Stanford University, Stanford, California, United States
COMET: A coordinated multi-layer runtime monitoring framework for human-supervised AI-enabled safety systems
Oct 21, 2026Niclas Flehmig, Mary Ann Lundteigen, Shen YinAI Safety AI Agent Monitoring
Department of Mechanical and Industrial Engineering, Norwegian University of Science and Technology, Trondheim, Norway · Department of Cybernetics Engineering, Norwegian University of Science and Technology, Trondheim, Norway
Correlated Failures of AI Approval Monitors
Oct 21, 2026Luca ZanoliniAI Agent Monitoring AI Agent Reliability
Independent researcher
RSI Control: Can AI Models Perform Covert Sabotage in the Self-Improvement Process?
Oct 20, 2026Kaiwen Zhou, Udari Madhushani SehwagRecursive Self-Improvement AI Control
Preparedness and Policy, Scale AI
Honeypots for AI Agent Insider Threats
Oct 19, 2026Jan Wehner, Jam KraprayoonAI Agent Security AI Agent Monitoring
AllPath: Layered Verification for a Deployed, Self-Revising LLM Trading Agent under Human-Gated and Autonomous Regimes
Oct 19, 2026Tian Zhang, Meng LiLLM Agent Verification LLM Agents
Independent Researcher USA
StopLoss: Teaching Agents to Predict Failure and Stop Early
Oct 18, 2026Jiarong Liang, Xueguang Ma, Dongfu Jiang +3Early Stopping AI Agent Monitoring
University of Waterloo
Mining the Self-Rewriting System: A Vision for Data Mining on AI-Native Infrastructure
Oct 18, 2026Mayank Darbari, Chiman Salavati, Liang Wu +5Distribution Shift AI Agent Monitoring
Nokia, Sunnyvale, California, USA · University of Connecticut, Storrs, Connecticut, USA
Red-Teaming Multi-Agent Attacks for Loss-of-Control Risks
Oct 16, 2026John Kitaoka, Thao Amelia Pham, Carlos Rafael Giudice +1AI Agent Safety Multi-Agent System Security
MATS Research · Redwood Research · Independent
LEASH: Uncertainty-Aware Weak-to-Strong Escalation and Oversight for Computer-Use Agents
Oct 15, 2026Qingni Wang, Changdae Oh, Yuzhe Yang +4Computer-Use Agents AI Agent Monitoring
Simular · University of Wisconsin–Madison
Agent Status Protocol (ASP): A Lightweight Specification for State Reporting in Multi-Agent Systems
Oct 15, 2026Diego Gosmar, Deborah A. Dahl, Emmett Coin +2Agent Communication Protocols AI Agent Monitoring
Open Voice Interoperability Initiative Linux Foundation AI & Data Turin, Italy · Conversational Technologies Open Voice Interoperability Initiative Linux Foundation AI & Data Plymouth Meeting, PA, USA · ejTalk, Inc. Open Voice Interoperability Initiative Linux Foundation AI & Data Belleville, Michigan, USA +2
Screwtape: Adversarial Transcript Editing for Monitor Red-Teaming
Oct 14, 2026Gosha Kozhevnikov, Tyler TracyLLM Red Teaming Automated Red Teaming
Redwood Research
ARIA: Runtime Inspection and Amendment for Multi-Agent Systems
Oct 14, 2026Bolin Shen, Yushun DongMulti-Agent LLM Systems Multi-Agent Systems
Florida State University
One Monitor Does Not Fit All: Formalism-Agnostic Runtime Verification for Agentic AI
Oct 14, 2026Angelo FerrandoAI Agent Monitoring Runtime Enforcement for AI Agents
University of Modena and Reggio Emilia, Modena, Italy
MCAIR: Monitoring and Correcting Agent Unfaithfulness via Internal Representations
Oct 14, 2026Yan Xu, Dandan Song, Luan Zhang +8AI Agent Monitoring LLM Agent Reliability
School of Computer Science and Technology, Beijing Institute of Technology, China
Limits of Statistical Assurance for Embodied and Agentic AI under Adaptive Adversaries
Oct 13, 2026Inderjeet SinghAI Assurance AI Safety Evaluation
Fujitsu Research of Europe, Slough, United Kingdom
Behavioural self-monitoring for autonomous earth observation agents: The attribute reliability monitor (ARM)
Oct 13, 2026Alejandro D. MousistAI Agent Monitoring AI Agent Reliability
Thales Alenia Space España C/ Einstein 7 (PTM) 28760 Tres Cantos, Madrid, Spain
HelpMeOut: A Neuropsychology-Grounded Benchmark for Error-Aware Monitoring of Household Task Execution
Oct 13, 2026Haoran Wang, Jingwei Huang, Anqing Duan +2AI Agent Monitoring AI Agent Benchmarks
University of Auckland, New Zealand · MBZUAI, United Arab Emirates
Dream2Discern: Learning from Hard Latent Variants for Runtime Failure Prediction in Multi-Agent Systems
Oct 12, 2026Boxuan Zhang, jianing Zhu, Xiaotian Han +1AI Agent Auditing AI Agent Monitoring
Rutgers University · The University of Texas at Austin
Revocation Is Not Complete at Receipt: An Executable Outcome Contract for Distributed Agent-Tool Gateways
Oct 11, 2026Lifei Liu, Haoran Yu, Danping ZhangRuntime Enforcement for AI Agents Access Control
Independent Researcher Seattle, WA, USA · Nanchang Hangkong University Nanchang, China
From Reactive Containment to Proactive Assurance: Lessons from OpenAI, Anthropic, and Google Agent Security Incidents
Oct 8, 2026Abbas RaftariAI Agent Security Runtime Enforcement for AI Agents
Walsh College, Troy, Michigan, USA

Oct 29, 2026J/K move · Enter open · S save

Pre-Crime for Agents, at the Last Second: Probing When a Recruited Agent Gives In

Leonidas Raghav

Independent

LLM Agent Security AI Agent Monitoring LLM Guardrails Linear Probing

Open paper AskPDF arXiv

When AI agents share channels and credentials, one agent can press another to break its operator's rule, a first step towards collusion. Output filters check only what an agent has already written. Can an overseer tell beforehand that an agent is about to give in? We show that a linear probe on the agent's activations can, in a controlled multi-agent setting. Across six scenarios, a peer presses a target for up to 20 rounds to reveal a secret, such as a deploy token; the target often refuses, then gives in. Read at each model call before its output is written, the probe picks out the one that leaks. On held-out Qwen3-32B episodes, at a threshold set to flag 5% of benign calls, it catches 82% of leaking calls. At a similar flag rate, a zero-shot decision-model monitor catches 26%. On gpt-oss-20b, a probe on its reasoning reaches AUROC 0.91. A bag-of-words classifier trained on the same calls comes close, so most of the gain comes from training on labels, not activations. The probe also flags permitted disclosures: it reads the coming disclosure more than the violation. The warning is short-range, and halting on any alarm would stop most benign episodes, which a second text check does not fix. A warning raised before generation could let an overseer stop an agent that is about to give in, ahead of others acting on its messages or recruitment spreading through a swarm.