{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:YODN42LEEMX4QMLRFBAL5RMCDM","short_pith_number":"pith:YODN42LE","schema_version":"1.0","canonical_sha256":"c386de6964232fc831712840bec5821b3f929131d28e021376af5f7e1377dbed","source":{"kind":"arxiv","id":"2502.15657","version":2},"attestation_state":"computed","paper":{"title":"Superintelligent Agents Pose Catastrophic Risks: Can Scientist AI Offer a Safer Path?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.AI","authors_text":"Adam Oberman, Damiano Fornasiere, David Williams-King, Jesse Richardson, Joumana Ghosn, Marc-Antoine Rondeau, Matt MacDermott, Michael Cohen, Oliver Richardson, Pierre-Luc St-Charles, Pietro Greiner, S\\\"oren Mindermann, Yoshua Bengio","submitted_at":"2025-02-21T18:28:36Z","abstract_excerpt":"The leading AI companies are increasingly focused on building generalist AI agents -- systems that can autonomously plan, act, and pursue goals across almost all tasks that humans can perform. Despite how useful these systems might be, unchecked AI agency poses significant risks to public safety and security, ranging from misuse by malicious actors to a potentially irreversible loss of human control. We discuss how these risks arise from current AI training methods. Indeed, various scenarios and experiments have demonstrated the possibility of AI agents engaging in deception or pursuing goals "},"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":"2502.15657","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2025-02-21T18:28:36Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"d40c9739e889110e756a7b5e70e2d612413fefac41051e8ff618dbe053be3223","abstract_canon_sha256":"ee419dadf3b5f385e0a07c6d5eba792d6eab5a760f16afe4b3ca7c0424e2f51e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:18:52.083683Z","signature_b64":"617gAnhZiSnbZPWJ6WxnujZidIPXjAH268z2PVI971HO3/sVBEpOkxWRIyP5zyJxhhczO4FR2S1ai1KEUcOmCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c386de6964232fc831712840bec5821b3f929131d28e021376af5f7e1377dbed","last_reissued_at":"2026-07-05T10:18:52.083186Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:18:52.083186Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Superintelligent Agents Pose Catastrophic Risks: Can Scientist AI Offer a Safer Path?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.AI","authors_text":"Adam Oberman, Damiano Fornasiere, David Williams-King, Jesse Richardson, Joumana Ghosn, Marc-Antoine Rondeau, Matt MacDermott, Michael Cohen, Oliver Richardson, Pierre-Luc St-Charles, Pietro Greiner, S\\\"oren Mindermann, Yoshua Bengio","submitted_at":"2025-02-21T18:28:36Z","abstract_excerpt":"The leading AI companies are increasingly focused on building generalist AI agents -- systems that can autonomously plan, act, and pursue goals across almost all tasks that humans can perform. Despite how useful these systems might be, unchecked AI agency poses significant risks to public safety and security, ranging from misuse by malicious actors to a potentially irreversible loss of human control. We discuss how these risks arise from current AI training methods. Indeed, various scenarios and experiments have demonstrated the possibility of AI agents engaging in deception or pursuing goals "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.15657","kind":"arxiv","version":2},"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/2502.15657/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":"2502.15657","created_at":"2026-07-05T10:18:52.083246+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.15657v2","created_at":"2026-07-05T10:18:52.083246+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.15657","created_at":"2026-07-05T10:18:52.083246+00:00"},{"alias_kind":"pith_short_12","alias_value":"YODN42LEEMX4","created_at":"2026-07-05T10:18:52.083246+00:00"},{"alias_kind":"pith_short_16","alias_value":"YODN42LEEMX4QMLR","created_at":"2026-07-05T10:18:52.083246+00:00"},{"alias_kind":"pith_short_8","alias_value":"YODN42LE","created_at":"2026-07-05T10:18:52.083246+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":16,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.25108","citing_title":"The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2606.26418","citing_title":"Unbiased Canonical Set-Valued Oracles Via Lattice Theory","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2606.08234","citing_title":"SciTrace: Trajectory-Aware Safety Reasoning for Scientific Discovery Agents","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2605.24197","citing_title":"A Sober Look at Agentic Misalignment in Automated Workflows","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2606.26418","citing_title":"Unbiased Canonical Set-Valued Oracles Via Lattice Theory","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2606.29657","citing_title":"Safety from Honesty in a Disinterested AI Predictor","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2504.01990","citing_title":"Advances and Challenges in Foundation Agents: From Brain-Inspired Intelligence to Evolutionary, Collaborative, and Safe Systems","ref_index":32,"is_internal_anchor":false},{"citing_arxiv_id":"2507.11810","citing_title":"Evolving Roles of LLMs in Scientific Innovation: Assistant, Collaborator, Scientist, and Evaluator","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2601.05905","citing_title":"Illusions of Confidence? Diagnosing LLM Truthfulness via Neighborhood Consistency","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2602.21843","citing_title":"The economic alignment problem of artificial intelligence","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11134","citing_title":"Spurious Correlation Learning in Preference Optimization: Mechanisms, Consequences, and Mitigation via Tie Training","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2503.09567","citing_title":"Towards Reasoning Era: A Survey of Long Chain-of-Thought for Reasoning Large Language Models","ref_index":46,"is_internal_anchor":false},{"citing_arxiv_id":"2605.03413","citing_title":"Learning to Theorize the World from Observation","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2604.07745","citing_title":"The Cartesian Cut in Agentic AI","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05227","citing_title":"Rethinking Data Curation in LLM Training: Online Reweighting Offers Better Generalization than Offline Methods","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2604.20805","citing_title":"Relative Principals, Pluralistic Alignment, and the Structural Value Alignment Problem","ref_index":5,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/YODN42LEEMX4QMLRFBAL5RMCDM","json":"https://pith.science/pith/YODN42LEEMX4QMLRFBAL5RMCDM.json","graph_json":"https://pith.science/api/pith-number/YODN42LEEMX4QMLRFBAL5RMCDM/graph.json","events_json":"https://pith.science/api/pith-number/YODN42LEEMX4QMLRFBAL5RMCDM/events.json","paper":"https://pith.science/paper/YODN42LE"},"agent_actions":{"view_html":"https://pith.science/pith/YODN42LEEMX4QMLRFBAL5RMCDM","download_json":"https://pith.science/pith/YODN42LEEMX4QMLRFBAL5RMCDM.json","view_paper":"https://pith.science/paper/YODN42LE","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.15657&json=true","fetch_graph":"https://pith.science/api/pith-number/YODN42LEEMX4QMLRFBAL5RMCDM/graph.json","fetch_events":"https://pith.science/api/pith-number/YODN42LEEMX4QMLRFBAL5RMCDM/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/YODN42LEEMX4QMLRFBAL5RMCDM/action/timestamp_anchor","attest_storage":"https://pith.science/pith/YODN42LEEMX4QMLRFBAL5RMCDM/action/storage_attestation","attest_author":"https://pith.science/pith/YODN42LEEMX4QMLRFBAL5RMCDM/action/author_attestation","sign_citation":"https://pith.science/pith/YODN42LEEMX4QMLRFBAL5RMCDM/action/citation_signature","submit_replication":"https://pith.science/pith/YODN42LEEMX4QMLRFBAL5RMCDM/action/replication_record"}},"created_at":"2026-07-05T10:18:52.083246+00:00","updated_at":"2026-07-05T10:18:52.083246+00:00"}