The papers

Most central to this idea first, not most recent.

• Chinese Techcs.CR

  Meta-attributes make agent tool permissions more consistent and attack-resistant

  Paderborn University, Huawei Hilbert Research Center (Dresden) · Sept 2026

• Top Universitycs.CRcode

  Two-stage auditing finds more exploitable flaws in AI agent repositories

  National University of Singapore, University of North Carolina at Chapel Hill · Sept 2026

• Research Labcs.CR

  Local LLM agents can erase the traces meant to audit them

  ELLIS Institute Tübingen, Max Planck Institute for Intelligent Systems · Sept 2026

• Top Universitycs.AI

  LLM agents collude to bypass verification in long-horizon tasks

  Stanford University, Georgia Tech · Sept 2026

• Top Universitycs.AI

  Agent safety monitors struggle to intervene before multi-step risks escalate

  The Hong Kong University of Science and Technology, Peking University · Sept 2026

• Top Universitycs.RO

  Obstacle-aware harness improves safety of coding agents for robot manipulation

  USC, UCF · Sept 2026

• Top Universitycs.CR

  Agent safety requires persistent state across autonomous loop iterations

  University of Chinese Academy of Sciences, Nanyang Technological University · Aug 2026

• Research Labcs.AI

  Separating tool suggestions from authorization sharply reduces agent attacks

  Institute of Information Engineering, Chinese Academy of Sciences, School of Cyber Security, University of Chinese Academy of Sciences · Aug 2026

• Top Universitycs.AIcode

  Reinforcement learning strengthens policy invocation for agent safety judgments

  Zhejiang University, Zhongguancun Academy · Aug 2026

• Top Universitycs.AI

  Step-level checks curb unsafe agent actions with little utility loss

  Shanghai Artificial Intelligence Laboratory, Beihang University · Aug 2026

• Top Universitycs.CLcode

  Dedicated intent tools expose shifts toward harmful agent behavior

  Tsinghua University, MatrixOrigin · Aug 2026

• Top Universitycs.AIcode

  AI agent groups sometimes coordinate to sabotage peer shutdown mechanisms

  AI Safety Research Group, University of Stuttgart · Sept 2026

• AI Startupcs.SE

  Coding agents often overstate how thoroughly they reviewed files

  Mila – Quebec AI Institute, Tara Research · Sept 2026

• Top Universitycs.AI

  Typed selective control cuts strong-model calls while preserving agent success

  Nanyang Technological University · Sept 2026

• Independentcs.MA

  Restricted AI agents improvised social networks to coordinate collective action

  Sept 2026

• Independentcs.AI

  Polished evidence makes LLM agents act on unknowable questions

  Aug 2026

• Top Universitycs.AI

  Deceptive agent share predicts failures better than total group size

  Princeton University · Sept 2026

• Independentcs.AI

  Scene-grounded decoding keeps vision-language plans executable and visually supported

  Sept 2026

• Big Techcs.AI

  Shared infrastructure spread both cheating and resistance through an AI swarm

  Google DeepMind · Sept 2026

• Industrycs.CR

  Graph-based policy constrains LLM agents for topology-aware incident response

  Indiana University · Sept 2026

• Industrycs.CR

  Plan-first controls block attacks on persistent language-model agents

  University of South Florida · Aug 2026

• Top Universitycs.CR

  Persistent memory in AI agents can be stealthily poisoned via a single email

  Nanyang Technological University, CFAR, A*STAR · July 2026