Adaptive Attacks on Trusted Monitors Subvert AI Control Protocols

• URL: http://arxiv.org/abs/2510.09462v1
• Date: Fri, 10 Oct 2025 15:12:44 GMT
• Title: Adaptive Attacks on Trusted Monitors Subvert AI Control Protocols
• Authors: Mikhail Terekhov, Alexander Panfilov, Daniil Dzenhaliou, Caglar Gulcehre, Maksym Andriushchenko, Ameya Prabhu, Jonas Geiping,
• Abstract summary: We study adaptive attacks by an untrusted model that knows the protocol and the monitor model.<n>We instantiate a simple adaptive attack vector by which the attacker embeds publicly known or zero-shot prompt injections in the model outputs.
• Score: 80.68060125494645
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: AI control protocols serve as a defense mechanism to stop untrusted LLM agents from causing harm in autonomous settings. Prior work treats this as a security problem, stress testing with exploits that use the deployment context to subtly complete harmful side tasks, such as backdoor insertion. In practice, most AI control protocols are fundamentally based on LLM monitors, which can become a central point of failure. We study adaptive attacks by an untrusted model that knows the protocol and the monitor model, which is plausible if the untrusted model was trained with a later knowledge cutoff or can search for this information autonomously. We instantiate a simple adaptive attack vector by which the attacker embeds publicly known or zero-shot prompt injections in the model outputs. Using this tactic, frontier models consistently evade diverse monitors and complete malicious tasks on two main AI control benchmarks. The attack works universally against current protocols that rely on a monitor. Furthermore, the recent Defer-to-Resample protocol even backfires, as its resampling amplifies the prompt injection and effectively reframes it as a best-of-$n$ attack. In general, adaptive attacks on monitor models represent a major blind spot in current control protocols and should become a standard component of evaluations for future AI control mechanisms.

Related papers

• Basic Legibility Protocols Improve Trusted Monitoring [1.8012533382373042]
  The AI Control research agenda aims to evaluate control protocols that prevent AI systems from taking harmful actions during deployment.<n>Because human oversight is expensive, one approach is trusted monitoring, where weaker, trusted models oversee stronger ones.<n>We introduce legibility protocols, which encourage the untrusted model to take actions that are easier for the monitor to understand.
  arXiv  Detail & Related papers  (2026-02-09T19:55:31Z)
• Bypassing AI Control Protocols via Agent-as-a-Proxy Attacks [12.356708678431183]
  Current defenses rely on monitoring protocols that jointly evaluate an agent's Chain-of-Thought (CoT) and tool-use actions to ensure alignment with user intent.<n>We demonstrate that these monitoring-based defenses can be bypassed via a novel Agent-as-a- Proxy attack.<n>Our findings suggest current monitoring-based agentic defenses are fundamentally fragile regardless of model scale.
  arXiv  Detail & Related papers  (2026-02-04T21:38:38Z)
• CaMeLs Can Use Computers Too: System-level Security for Computer Use Agents [60.98294016925157]
  AI agents are vulnerable to prompt injection attacks, where malicious content hijacks agent behavior to steal credentials or cause financial loss.<n>We introduce Single-Shot Planning for CUAs, where a trusted planner generates a complete execution graph with conditional branches before any observation of potentially malicious content.<n>Although this architectural isolation successfully prevents instruction injections, we show that additional measures are needed to prevent Branch Steering attacks.
  arXiv  Detail & Related papers  (2026-01-14T23:06:35Z)
• Breaking and Fixing Defenses Against Control-Flow Hijacking in Multi-Agent Systems [7.558938027515112]
  Control-flow hijacking attacks manipulate orchestration mechanisms in multi-agent systems into performing unsafe actions.<n>We propose, implement, and evaluate ControlValve, a new defense inspired by the principles of control-flow integrity and least privilege.
  arXiv  Detail & Related papers  (2025-10-20T08:02:51Z)
• TabVLA: Targeted Backdoor Attacks on Vision-Language-Action Models [63.51290426425441]
  A backdoored VLA agent can be covertly triggered by a pre-injected backdoor to execute adversarial actions.<n>We study targeted backdoor attacks on VLA models and introduce TabVLA, a novel framework that enables such attacks via black-box fine-tuning.<n>Our work highlights the vulnerability of VLA models to targeted backdoor manipulation and underscores the need for more advanced defenses.
  arXiv  Detail & Related papers  (2025-10-13T02:45:48Z)
• Malice in Agentland: Down the Rabbit Hole of Backdoors in the AI Supply Chain [82.98626829232899]
  Fine-tuning AI agents on data from their own interactions introduces a critical security vulnerability within the AI supply chain.<n>We show that adversaries can easily poison the data collection pipeline to embed hard-to-detect backdoors.
  arXiv  Detail & Related papers  (2025-10-03T12:47:21Z)
• BlindGuard: Safeguarding LLM-based Multi-Agent Systems under Unknown Attacks [58.959622170433725]
  BlindGuard is an unsupervised defense method that learns without requiring any attack-specific labels or prior knowledge of malicious behaviors.<n>We show that BlindGuard effectively detects diverse attack types (i.e., prompt injection, memory poisoning, and tool attack) across multi-agent systems.
  arXiv  Detail & Related papers  (2025-08-11T16:04:47Z)
• CoT Red-Handed: Stress Testing Chain-of-Thought Monitoring [3.6284577335311563]
  Chain-of-Thought (CoT) monitoring improves detection by up to 27 percentage points in scenarios where action-only monitoring fails to reliably identify sabotage.<n>CoT traces can also contain misleading rationalizations that deceive the monitor, reducing performance in more obvious sabotage cases.<n>This hybrid monitor consistently outperforms both CoT and action-only monitors across all tested models and tasks, with detection rates over four times higher than action-only monitoring for subtle deception scenarios.
  arXiv  Detail & Related papers  (2025-05-29T15:47:36Z)
• CANTXSec: A Deterministic Intrusion Detection and Prevention System for CAN Bus Monitoring ECU Activations [53.036288487863786]
  We propose CANTXSec, the first deterministic Intrusion Detection and Prevention system based on physical ECU activations.<n>It detects and prevents classical attacks in the CAN bus, while detecting advanced attacks that have been less investigated in the literature.<n>We prove the effectiveness of our solution on a physical testbed, where we achieve 100% detection accuracy in both classes of attacks while preventing 100% of FIAs.
  arXiv  Detail & Related papers  (2025-05-14T13:37:07Z)
• LlamaFirewall: An open source guardrail system for building secure AI agents [0.5603362829699733]
  Large language models (LLMs) have evolved from simple chatbots into autonomous agents capable of performing complex tasks.<n>Given the higher stakes and the absence of deterministic solutions to mitigate these risks, there is a critical need for a real-time guardrail monitor.<n>We introduce LlamaFirewall, an open-source security focused guardrail framework.
  arXiv  Detail & Related papers  (2025-05-06T14:34:21Z)
• Ctrl-Z: Controlling AI Agents via Resampling [2.505902477669827]
  We construct BashBench, a dataset of 257 challenging multi-step system administration tasks.<n>We evaluate whether various safety measures can prevent an adversarially constructed AI agent from covertly downloading and executing malicious code.
  arXiv  Detail & Related papers  (2025-04-14T16:22:11Z)
• MELON: Provable Defense Against Indirect Prompt Injection Attacks in AI Agents [60.30753230776882]
  LLM agents are vulnerable to indirect prompt injection (IPI) attacks, where malicious tasks embedded in tool-retrieved information can redirect the agent to take unauthorized actions.<n>We present MELON, a novel IPI defense that detects attacks by re-executing the agent's trajectory with a masked user prompt modified through a masking function.
  arXiv  Detail & Related papers  (2025-02-07T18:57:49Z)
• Certifiers Make Neural Networks Vulnerable to Availability Attacks [70.69104148250614]
  We show for the first time that fallback strategies can be deliberately triggered by an adversary. In addition to naturally occurring abstains for some inputs and perturbations, the adversary can use training-time attacks to deliberately trigger the fallback. We design two novel availability attacks, which show the practical relevance of these threats.
  arXiv  Detail & Related papers  (2021-08-25T15:49:10Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.