{
  "id": "secret-collusion",
  "title": "Secret Collusion among AI Agents: Multi-Agent Deception via Steganography",
  "authors": [
    "Sumeet Ramesh Motwani",
    "Mikhail Baranchuk",
    "Martin Strohmeier",
    "Vijay Bolina",
    "Philip H.S. Torr",
    "Lewis Hammond",
    "Christian Schroeder de Witt"
  ],
  "author_parts": [
    {
      "given": "Sumeet Ramesh",
      "family": "Motwani"
    },
    {
      "given": "Mikhail",
      "family": "Baranchuk"
    },
    {
      "given": "Martin",
      "family": "Strohmeier"
    },
    {
      "given": "Vijay",
      "family": "Bolina"
    },
    {
      "given": "Philip H.S.",
      "family": "Torr"
    },
    {
      "given": "Lewis",
      "family": "Hammond"
    },
    {
      "given": "Christian",
      "family": "Schroeder de Witt"
    }
  ],
  "year": 2024,
  "venue_date": "2024",
  "venue": "Advances in Neural Information Processing Systems",
  "venue_short": "NeurIPS 2024",
  "publication_type": "conference",
  "publisher": null,
  "volume": "37",
  "doi": "10.52202/079017-2336",
  "arxiv_id": "2402.07510",
  "relation": "related",
  "categories": [
    "AI agents",
    "Covert communication",
    "Evaluation"
  ],
  "summary": "Formalizes secret collusion among generative AI agents and develops capability evaluations for covert communication and coordination. It studies individual and multi-agent settings alongside the assumptions and limitations of monitoring and other mitigations.",
  "relationship": "Section 2 identifies Ziegler et al. 2019 as an early AI-generated steganography approach using arithmetic coding; the paper studies the broader agent-security setting rather than extending the STEGASURAS implementation.",
  "limitation": "Its evaluated models show limited advanced steganographic capabilities, and the authors call for more complex multi-agent and real-world evaluations; the results are not evidence that every agent autonomously uses covert channels.",
  "sources": [
    {
      "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/861f7dad098aec1c3560fb7add468d41-Abstract-Conference.html",
      "evidence": "Published title, complete authors, NeurIPS 2024 volume 37 and DOI."
    },
    {
      "url": "https://arxiv.org/abs/2402.07510",
      "evidence": "First submission February 12, 2024 and subsequent version dates."
    },
    {
      "url": "https://arxiv.org/html/2402.07510v3",
      "evidence": "Conference-era full text; section 2 links original arithmetic work, sections 5–6 describe evaluations, section 8 states current capability limits and evaluation needs."
    },
    {
      "url": "https://arxiv.org/abs/2402.07510v1",
      "evidence": "Version-specific record verifies original title and February 12, 2024 submission date."
    }
  ],
  "arxiv_url": "https://arxiv.org/abs/2402.07510",
  "earliest_title": "Secret Collusion Among Generative AI Agents",
  "earliest_date": "2024-02-12",
  "earliest_date_label": "First arXiv submission",
  "url": "https://proceedings.neurips.cc/paper_files/paper/2024/hash/861f7dad098aec1c3560fb7add468d41-Abstract-Conference.html",
  "pdf": "https://proceedings.nips.cc/paper_files/paper/2024/file/861f7dad098aec1c3560fb7add468d41-Paper-Conference.pdf",
  "code": null,
  "category": "agents",
  "contribution": "Formalizes secret collusion among generative AI agents and develops capability evaluations for covert communication and coordination. It studies individual and multi-agent settings alongside the assumptions and limitations of monitoring and other mitigations.",
  "canonical_url": "https://steganography.live/reading-list/papers/secret-collusion/"
}