{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:RJQ7OPHVBB7DVOD5VQNCLVJ2ZN","short_pith_number":"pith:RJQ7OPHV","schema_version":"1.0","canonical_sha256":"8a61f73cf5087e3ab87dac1a25d53acb56bc9c98f5ce618ce406e239a9f74817","source":{"kind":"arxiv","id":"2405.01576","version":1},"attestation_state":"computed","paper":{"title":"Uncovering Deceptive Tendencies in Language Models: A Simulated Company AI Assistant","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Evan Hubinger, Olli J\\\"arviniemi","submitted_at":"2024-04-25T17:29:53Z","abstract_excerpt":"We study the tendency of AI systems to deceive by constructing a realistic simulation setting of a company AI assistant. The simulated company employees provide tasks for the assistant to complete, these tasks spanning writing assistance, information retrieval and programming. We then introduce situations where the model might be inclined to behave deceptively, while taking care to not instruct or otherwise pressure the model to do so. Across different scenarios, we find that Claude 3 Opus\n  1) complies with a task of mass-generating comments to influence public perception of the company, late"},"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":"2405.01576","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-04-25T17:29:53Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"900d0815bbec065520f6caeff199fdda4568da7a9fcc96092243ba6a14d5d48e","abstract_canon_sha256":"9290bf8c52d3b0a5d2dbee9702b76c28576dd2f8ec1d8fc1e40e6f4bc694676c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:14:58.257915Z","signature_b64":"QYLtqxPUdODZHqHHy+QarR1qkmNx1XezwEyeCACpjo8KKpst4+YocuPVv8Ntp9PEezRZqx7pbe6AqG2U6SnXAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8a61f73cf5087e3ab87dac1a25d53acb56bc9c98f5ce618ce406e239a9f74817","last_reissued_at":"2026-07-05T08:14:58.257486Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:14:58.257486Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Uncovering Deceptive Tendencies in Language Models: A Simulated Company AI Assistant","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Evan Hubinger, Olli J\\\"arviniemi","submitted_at":"2024-04-25T17:29:53Z","abstract_excerpt":"We study the tendency of AI systems to deceive by constructing a realistic simulation setting of a company AI assistant. The simulated company employees provide tasks for the assistant to complete, these tasks spanning writing assistance, information retrieval and programming. We then introduce situations where the model might be inclined to behave deceptively, while taking care to not instruct or otherwise pressure the model to do so. Across different scenarios, we find that Claude 3 Opus\n  1) complies with a task of mass-generating comments to influence public perception of the company, late"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.01576","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/2405.01576/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":"2405.01576","created_at":"2026-07-05T08:14:58.257544+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.01576v1","created_at":"2026-07-05T08:14:58.257544+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.01576","created_at":"2026-07-05T08:14:58.257544+00:00"},{"alias_kind":"pith_short_12","alias_value":"RJQ7OPHVBB7D","created_at":"2026-07-05T08:14:58.257544+00:00"},{"alias_kind":"pith_short_16","alias_value":"RJQ7OPHVBB7DVOD5","created_at":"2026-07-05T08:14:58.257544+00:00"},{"alias_kind":"pith_short_8","alias_value":"RJQ7OPHV","created_at":"2026-07-05T08:14:58.257544+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.07157","citing_title":"Think Fast: Estimating No-CoT Task-Completion Time Horizons of Frontier AI Models","ref_index":77,"is_internal_anchor":false},{"citing_arxiv_id":"2606.07157","citing_title":"Think Fast: Estimating No-CoT Task-Completion Time Horizons of Frontier AI Models","ref_index":77,"is_internal_anchor":false},{"citing_arxiv_id":"2412.16720","citing_title":"OpenAI o1 System Card","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2412.04984","citing_title":"Frontier Models are Capable of In-context Scheming","ref_index":19,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/RJQ7OPHVBB7DVOD5VQNCLVJ2ZN","json":"https://pith.science/pith/RJQ7OPHVBB7DVOD5VQNCLVJ2ZN.json","graph_json":"https://pith.science/api/pith-number/RJQ7OPHVBB7DVOD5VQNCLVJ2ZN/graph.json","events_json":"https://pith.science/api/pith-number/RJQ7OPHVBB7DVOD5VQNCLVJ2ZN/events.json","paper":"https://pith.science/paper/RJQ7OPHV"},"agent_actions":{"view_html":"https://pith.science/pith/RJQ7OPHVBB7DVOD5VQNCLVJ2ZN","download_json":"https://pith.science/pith/RJQ7OPHVBB7DVOD5VQNCLVJ2ZN.json","view_paper":"https://pith.science/paper/RJQ7OPHV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.01576&json=true","fetch_graph":"https://pith.science/api/pith-number/RJQ7OPHVBB7DVOD5VQNCLVJ2ZN/graph.json","fetch_events":"https://pith.science/api/pith-number/RJQ7OPHVBB7DVOD5VQNCLVJ2ZN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RJQ7OPHVBB7DVOD5VQNCLVJ2ZN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RJQ7OPHVBB7DVOD5VQNCLVJ2ZN/action/storage_attestation","attest_author":"https://pith.science/pith/RJQ7OPHVBB7DVOD5VQNCLVJ2ZN/action/author_attestation","sign_citation":"https://pith.science/pith/RJQ7OPHVBB7DVOD5VQNCLVJ2ZN/action/citation_signature","submit_replication":"https://pith.science/pith/RJQ7OPHVBB7DVOD5VQNCLVJ2ZN/action/replication_record"}},"created_at":"2026-07-05T08:14:58.257544+00:00","updated_at":"2026-07-05T08:14:58.257544+00:00"}