{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:MTLFQTOUX3CUPTV267AX46CBCH","short_pith_number":"pith:MTLFQTOU","schema_version":"1.0","canonical_sha256":"64d6584dd4bec547cebaf7c17e784111c60327d2693c4e9504bd82fb1285063f","source":{"kind":"arxiv","id":"2502.14143","version":1},"attestation_state":"computed","paper":{"title":"Multi-Agent Risks from Advanced AI","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CY","cs.ET","cs.LG"],"primary_cat":"cs.MA","authors_text":"Adam Gleave, Akbir Khan, Alan Chan, Alexander Meinke, Anka Reuel, Atoosa Kasirzadeh, Carson Ezell, Caspar Oesterheld, Chandler Smith, Christian Schroeder de Witt, David C. Parkes, Edward Hughes, Esben Kran, Euan McLean, Georgios Piliouras, Gillian Hadfield, Iason Gabriel, Igor Krawczuk, Iyad Rahwan, Jakob Foerster, Jan Kulveit, Jason Hoelscher-Obermaier, Jesse Clifton, Joel Lehman, Joel Z. Leibo, Kate Larson, Lewis Hammond, Matija Franklin, Max Lamparth, Michael Dennis, Michael Wellman, Nika Haghtalab, Niklas Lauffer, Nisarg Shah, Paolo Bova, Quentin Feuillade-Montixi, S\\'ebastien Krier, Sumeet Motwani, The Anh Han, Theodor Cimpeanu, Tom\\'a\\v{s} Gaven\\v{c}iak, Vincent Conitzer, Vojt\\v{e}ch Kova\\v{r}\\'ik, Wolfram Barfuss","submitted_at":"2025-02-19T23:03:21Z","abstract_excerpt":"The rapid development of advanced AI agents and the imminent deployment of many instances of these agents will give rise to multi-agent systems of unprecedented complexity. These systems pose novel and under-explored risks. In this report, we provide a structured taxonomy of these risks by identifying three key failure modes (miscoordination, conflict, and collusion) based on agents' incentives, as well as seven key risk factors (information asymmetries, network effects, selection pressures, destabilising dynamics, commitment problems, emergent agency, and multi-agent security) that can underp"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2502.14143","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.MA","submitted_at":"2025-02-19T23:03:21Z","cross_cats_sorted":["cs.AI","cs.CY","cs.ET","cs.LG"],"title_canon_sha256":"7a0a9d0364f1b018d7dd4a6ce7d02978063a508ff217c65862e41fc88cf64fb3","abstract_canon_sha256":"cb49c66f82105f2d6b7fe874c1ad9ed82fa539af1bae5bc2c52e052fe5192664"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:17:21.510859Z","signature_b64":"m4hDa/gf/FXkWniVgOZZ9/BjTUbyGkf/NWcsy57I/8PXhC/oKtZtJJievtGcjlcG67Fa+dN2i0FFWo+OTDCkCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"64d6584dd4bec547cebaf7c17e784111c60327d2693c4e9504bd82fb1285063f","last_reissued_at":"2026-07-05T10:17:21.510323Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:17:21.510323Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Multi-Agent Risks from Advanced AI","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CY","cs.ET","cs.LG"],"primary_cat":"cs.MA","authors_text":"Adam Gleave, Akbir Khan, Alan Chan, Alexander Meinke, Anka Reuel, Atoosa Kasirzadeh, Carson Ezell, Caspar Oesterheld, Chandler Smith, Christian Schroeder de Witt, David C. Parkes, Edward Hughes, Esben Kran, Euan McLean, Georgios Piliouras, Gillian Hadfield, Iason Gabriel, Igor Krawczuk, Iyad Rahwan, Jakob Foerster, Jan Kulveit, Jason Hoelscher-Obermaier, Jesse Clifton, Joel Lehman, Joel Z. Leibo, Kate Larson, Lewis Hammond, Matija Franklin, Max Lamparth, Michael Dennis, Michael Wellman, Nika Haghtalab, Niklas Lauffer, Nisarg Shah, Paolo Bova, Quentin Feuillade-Montixi, S\\'ebastien Krier, Sumeet Motwani, The Anh Han, Theodor Cimpeanu, Tom\\'a\\v{s} Gaven\\v{c}iak, Vincent Conitzer, Vojt\\v{e}ch Kova\\v{r}\\'ik, Wolfram Barfuss","submitted_at":"2025-02-19T23:03:21Z","abstract_excerpt":"The rapid development of advanced AI agents and the imminent deployment of many instances of these agents will give rise to multi-agent systems of unprecedented complexity. These systems pose novel and under-explored risks. In this report, we provide a structured taxonomy of these risks by identifying three key failure modes (miscoordination, conflict, and collusion) based on agents' incentives, as well as seven key risk factors (information asymmetries, network effects, selection pressures, destabilising dynamics, commitment problems, emergent agency, and multi-agent security) that can underp"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.14143","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2502.14143/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2502.14143","created_at":"2026-07-05T10:17:21.510385+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.14143v1","created_at":"2026-07-05T10:17:21.510385+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.14143","created_at":"2026-07-05T10:17:21.510385+00:00"},{"alias_kind":"pith_short_12","alias_value":"MTLFQTOUX3CU","created_at":"2026-07-05T10:17:21.510385+00:00"},{"alias_kind":"pith_short_16","alias_value":"MTLFQTOUX3CUPTV2","created_at":"2026-07-05T10:17:21.510385+00:00"},{"alias_kind":"pith_short_8","alias_value":"MTLFQTOU","created_at":"2026-07-05T10:17:21.510385+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":29,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.22495","citing_title":"Grounded Scaling: Why Agentic AI Needs Deterministic Environments","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2606.12918","citing_title":"MAStrike: Shapley-Guided Collusive Red-Teaming on Multi-Agent Systems","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2606.11533","citing_title":"AI Researchers Must Help Lead Arms Control to Mitigate Military AI Risks","ref_index":91,"is_internal_anchor":false},{"citing_arxiv_id":"2606.10711","citing_title":"The Agentic Web Requires New Normative Infrastructure","ref_index":240,"is_internal_anchor":false},{"citing_arxiv_id":"2606.09414","citing_title":"AI Assurance in UK Defence: Challenges in Operationalising JSP 936","ref_index":84,"is_internal_anchor":false},{"citing_arxiv_id":"2606.09931","citing_title":"A Note on the Strategic Confinement Problem","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2606.08340","citing_title":"Benchmarking Open-Ended Multi-Agent Coordination in Language Agents","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2607.00913","citing_title":"Two AI Metrics Diverged: Will it Make All the Difference?","ref_index":70,"is_internal_anchor":false},{"citing_arxiv_id":"2606.00047","citing_title":"Comprehensive AI governance requires addressing non-model gains","ref_index":45,"is_internal_anchor":false},{"citing_arxiv_id":"2605.30169","citing_title":"Dissociative Identity: Language Model Agents Lack Grounding for Reputation Mechanisms","ref_index":58,"is_internal_anchor":false},{"citing_arxiv_id":"2606.10711","citing_title":"The Agentic Web Requires New Normative Infrastructure","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2606.21296","citing_title":"Discriminatory Compliance: How LLMs Answer Queries from Protected Groups","ref_index":38,"is_internal_anchor":false},{"citing_arxiv_id":"2605.23414","citing_title":"When Planning Fails Despite Correct Execution: On Epistemic Calibration for LLM-Based Multi-Agent Systems","ref_index":128,"is_internal_anchor":false},{"citing_arxiv_id":"2601.22292","citing_title":"Learning Incentive Structures for Cooperative Resilience in Multi-Agent Systems under Social Dilemmas","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2605.16278","citing_title":"Keeping an Eye on AI: A Framework for Effective Human Oversight of AI Systems","ref_index":72,"is_internal_anchor":false},{"citing_arxiv_id":"2605.20704","citing_title":"Heartbeat-Bound Hierarchical Credentials: Cryptographic Revocation for AI Agent Swarms","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2605.16282","citing_title":"Taxonomy and Consistency Analysis of Safety Benchmarks for AI Agents","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2510.12826","citing_title":"Scheming Ability in LLM-to-LLM Strategic Interactions","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2510.14133","citing_title":"Formalizing the Safety, Security, and Functional Properties of Agentic AI Systems","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2601.03846","citing_title":"When Numbers Start Talking: Implicit Numerical Coordination Among LLM-Based Agents","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2603.27771","citing_title":"Emergent Social Intelligence Risks in Generative Multi-Agent Systems","ref_index":47,"is_internal_anchor":false},{"citing_arxiv_id":"2604.01151","citing_title":"Detecting Multi-Agent Collusion Through Multi-Agent Interpretability","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10481","citing_title":"Safe Multi-Agent Behavior Must Be Maintained, Not Merely Asserted: Constraint Drift in LLM-Based Multi-Agent Systems","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2503.13657","citing_title":"Why Do Multi-Agent LLM Systems Fail?","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2604.24572","citing_title":"FastOMOP: A Foundational Architecture for Reliable Agentic Real-World Evidence Generation on OMOP CDM data","ref_index":17,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/MTLFQTOUX3CUPTV267AX46CBCH","json":"https://pith.science/pith/MTLFQTOUX3CUPTV267AX46CBCH.json","graph_json":"https://pith.science/api/pith-number/MTLFQTOUX3CUPTV267AX46CBCH/graph.json","events_json":"https://pith.science/api/pith-number/MTLFQTOUX3CUPTV267AX46CBCH/events.json","paper":"https://pith.science/paper/MTLFQTOU"},"agent_actions":{"view_html":"https://pith.science/pith/MTLFQTOUX3CUPTV267AX46CBCH","download_json":"https://pith.science/pith/MTLFQTOUX3CUPTV267AX46CBCH.json","view_paper":"https://pith.science/paper/MTLFQTOU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.14143&json=true","fetch_graph":"https://pith.science/api/pith-number/MTLFQTOUX3CUPTV267AX46CBCH/graph.json","fetch_events":"https://pith.science/api/pith-number/MTLFQTOUX3CUPTV267AX46CBCH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MTLFQTOUX3CUPTV267AX46CBCH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MTLFQTOUX3CUPTV267AX46CBCH/action/storage_attestation","attest_author":"https://pith.science/pith/MTLFQTOUX3CUPTV267AX46CBCH/action/author_attestation","sign_citation":"https://pith.science/pith/MTLFQTOUX3CUPTV267AX46CBCH/action/citation_signature","submit_replication":"https://pith.science/pith/MTLFQTOUX3CUPTV267AX46CBCH/action/replication_record"}},"created_at":"2026-07-05T10:17:21.510385+00:00","updated_at":"2026-07-05T10:17:21.510385+00:00"}