Distributed Attacks in Persistent-State AI Control

• URL: http://arxiv.org/abs/2607.02514v2
• Date: Wed, 08 Jul 2026 20:25:40 GMT
• Title: Distributed Attacks in Persistent-State AI Control
• Abstract summary: Iterative VibeCoding is the study of safely deploying capable but potentially untrusted AI.<n>Evasion remains high across state-of-the-art monitor models.<n>We introduce a stateful link-tracker monitor that tracks suspicious buildup across PRs.
• Score: 0.8411355178908656
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: As AI coding agents become more autonomous, they increasingly ship code iteratively, with the codebase persisting across sessions. This persistence creates a new attack surface: a misaligned or prompt-injected agent can distribute attacks across pull requests (PRs) and time its payload for the PR with the best natural cover. To study the resulting dynamics, we introduce Iterative VibeCoding, a setting for AI control, the study of safely deploying capable but potentially untrusted AI. In Iterative VibeCoding, a coding agent builds software over a sequence of PRs in a persistent codebase while pursuing a covert side task. Our benchmark includes two task families: CLI tools and Flask web services, across 20 total task variations. We use Claude Sonnet 4.5 as the attack agent and GPT-4o as the monitor. We compare gradual attacks, which distribute the side task across PRs, against non-gradual attacks concentrated in a single PR. No single monitor is robust to both: which strategy evades best (success while evading the monitor) depends on the monitor type, so a defender cannot close off both gradual and non-gradual attacks with any one monitor. High evasion (>= 65%) generalizes across model attack agent backends (Sonnet 4.5, Gemini 3.1 Pro, Kimi K2.5), confirming this is a property of the persistent-state attack surface rather than a single model's capability. Evasion also remains high across state-of-the-art monitor models and the gap between gradual and non-gradual evasion widens for more capable models. We introduce a stateful link-tracker monitor that tracks suspicious buildup across PRs. On both task families, it detects gradual attacks substantially better than diff monitors that merely see more accumulated history. Combining this stronger monitor with trajectory monitors in a four-monitor ensemble reduces gradual-attack evasion from 93% under the weakest standard diff monitor to 47%.

Related papers

• ResearchArena: Evaluating Sabotage and Monitoring in Automated AI R&D [11.818511808737037]
  We evaluate AI control for automated AI R&D with ResearchArena.<n>We pair each main task with two kinds of hidden side task.<n>We find that sabotage hidden in the training data is the hardest to catch.
  arXiv  Detail & Related papers  (2026-07-21T17:41:12Z)
• Multi-Agent AI Control: Distributed Attacks Hamper Per-Instance Monitors [4.246803713067298]
  We study multi-agent AI control, formalising distributed attacks in which several agents jointly aim for a malicious goal.<n>We develop FakeLab: a synthetic AI-lab (9 services, 86 benign tasks, 4 attack objectives)<n>Our central finding is the fragmentation effect: as more agents coordinate to attack, per-agent monitoring becomes less likely to catch any of the attackers.
  arXiv  Detail & Related papers  (2026-07-08T13:03:25Z)
• Stateful Online Monitoring Catches Distributed Agent Attacks [63.00394432922707]
  We develop an online stateful monitor that uses real-time clustering to collect weak suspiciousness signals across many agent transcripts.<n>Our results point toward a new class of safety monitors which reason over groups of users rather than isolated transcripts.
  arXiv  Detail & Related papers  (2026-05-29T17:57:00Z)
• SLEIGHT-Bench: A Benchmark of Evasion Attacks Against Agent Monitors [3.989565392866092]
  We present SLEIGHT-Bench (Subtle Low-itEration Insight-Guided Harmful Transcripts), a benchmark of synthetic transcripts containing 40 attacks across 11 categories.<n>We find that 20 of 40 attacks are never caught by an Opus 4.6 monitor with extended thinking across 10 trials at a 1% false-positive rate, while the overall catch rate is 32%.
  arXiv  Detail & Related papers  (2026-05-15T20:46:46Z)
• MonitoringBench: Semi-Automated Red-Teaming for Agent Monitoring [0.0]
  We introduce a red-teaming methodology that exposes harder-to-catch attacks for monitoring agents.<n>We find that frontier monitors often detect suspicious actions but fall for persuasion or fail to calibrate suspiciousness scores appropriately.<n>MonitoringBench is a benchmark of 2,644 attack trajectories for evaluating monitor capabilities and failure modes.
  arXiv  Detail & Related papers  (2026-05-10T18:04:31Z)
• WebAgentGuard: A Reasoning-Driven Guard Model for Detecting Prompt Injection Attacks in Web Agents [117.65855863464863]
  Web agents are highly vulnerable to prompt injection attacks.<n>Existing defenses, including system prompt defenses and direct fine-tuning of agents, have shown limited effectiveness.<n>We introduce WebAgentGuard, a reasoning-driven, multimodal guard model for prompt injection detection.
  arXiv  Detail & Related papers  (2026-04-14T04:50:35Z)
• Async Control: Stress-testing Asynchronous Control Measures for LLM Agents [2.7146936326590425]
  We investigate asynchronous monitoring, in which a monitoring system reviews agent actions after the fact.<n>Unlike synchronous monitoring, this approach does not impose runtime latency, while still attempting to disrupt attacks before irreversible harm occurs.<n>We treat monitor development as an adversarial game between a blue team (who design monitors) and a red team (who create sabotaging agents)
  arXiv  Detail & Related papers  (2025-12-15T16:56:28Z)
• Adaptive Attacks on Trusted Monitors Subvert AI Control Protocols [80.68060125494645]
  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.
  arXiv  Detail & Related papers  (2025-10-10T15:12:44Z)
• 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)
• Poison Once, Control Anywhere: Clean-Text Visual Backdoors in VLM-based Mobile Agents [54.35629963816521]
  This work introduces VIBMA, the first clean-text backdoor attack targeting VLM-based mobile agents.<n>The attack injects malicious behaviors into the model by modifying only the visual input.<n>We show that our attack achieves high success rates while preserving clean-task behavior.
  arXiv  Detail & Related papers  (2025-06-16T08:09:32Z)
• Monitoring Reasoning Models for Misbehavior and the Risks of Promoting Obfuscation [56.102976602468615]
  We show that we can monitor a frontier reasoning model, such as OpenAI o3-mini, for reward hacking in agentic coding environments.<n>We find that with too much optimization, agents learn obfuscated reward hacking, hiding their intent within the chain-of-thought.
  arXiv  Detail & Related papers  (2025-03-14T23:50:34Z)
• Dissecting Adversarial Robustness of Multimodal LM Agents [70.2077308846307]
  We manually create 200 targeted adversarial tasks and evaluation scripts in a realistic threat model on top of VisualWebArena.<n>We find that we can successfully break latest agents that use black-box frontier LMs, including those that perform reflection and tree search.<n>We also use ARE to rigorously evaluate how the robustness changes as new components are added.
  arXiv  Detail & Related papers  (2024-06-18T17:32:48Z)

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.