Metraipolis › archive › dc_25495bbaa4ed

Covert Assistance: Helpful LLM Agents Evade Oversight in Multi-Agent Systems — arXiv:2609.39050

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.39050
License (as archived)
CC0-1.0
Archived
2026-10-07T13:14:23.661Z
SHA-256 of the archived text
4fbbe40b240c6f3cfe2503f49ad9d51d92abe910124a0cf3dd5d9759369cb087
Stable URL of this page
https://metraipolis.com/documento/dc_25495bbaa4ed

How to cite

Covert Assistance: Helpful LLM Agents Evade Oversight in Multi-Agent Systems — arXiv:2609.39050 — arXiv (en), https://arxiv.org/abs/2609.39050 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_25495bbaa4ed (Metraipolis archive dc_25495bbaa4ed, sha256 4fbbe40b240c6f3cfe2503f49ad9d51d92abe910124a0cf3dd5d9759369cb087)
BibTeX
@misc{dc25495bbaa4ed,
  title = {Covert Assistance: Helpful LLM Agents Evade Oversight in Multi-Agent Systems — arXiv:2609.39050},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.39050},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_25495bbaa4ed; sha256 4fbbe40b240c6f3cfe2503f49ad9d51d92abe910124a0cf3dd5d9759369cb087},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.39050. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

As multi-agent systems enter high-stakes domains, the possibility that agents may circumvent safety boundaries is a growing concern. Prior work has examined this risk primarily in adversarial settings, where agents are instructed or rewarded to communicate covertly and evade oversight. We show that benign agents can cross the same boundaries without adversarial incentives. We emulate a software-engineering workflow in which a planner represents a company hiring an external developer. The planner writes requirements and holds a company credential it is instructed not to disclose to the developer; a monitor screens their exchanges. Seven of nine tested frontier models disguise the credential in their requirements to help the developer recover it while evading the monitor, even after completing their assigned objective. For example, across 6,000 episodes with DeepSeek-V4-Pro, the planner attempts concealment in 16.9%; in 0.9%, the credential evades the monitor and is recovered and used by the developer. Despite the low per-episode rate, the risk compounds across repeated exchanges: at 0.9%, 105 independent episodes carry a 61.3% chance of at least one breach. Our analysis suggests that the planner misinterprets disguised disclosure as helping the developer and reads the nondisclosure rule as banning plaintext, not character codes or riddles. We call this behavior covert assistance. Monitoring favors concealment, while concealment gives way to direct disclosure when the recipient is presented as a human rather than an AI agent, so leakage persists. Giving the monitor the credential improves detection but requires trusting it with the secret. These risks, in models already used for software engineering, challenge oversight to distinguish authorized cooperation from task-advancing assistance that crosses safety boundaries.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-10-01 in cs.CL).
Categorie arXiv: Cryptography and Security (cs.CR) ; Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2609.39050 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-07T12:44:51.956Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-10-01?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).