{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:HSM7VA7NQLVBZ672CZ6FEECCAK","short_pith_number":"pith:HSM7VA7N","schema_version":"1.0","canonical_sha256":"3c99fa83ed82ea1cfbfa167c52104202ad164850c61ea158d7b6f61ef437e8af","source":{"kind":"arxiv","id":"2411.03336","version":2},"attestation_state":"computed","paper":{"title":"Towards evaluations-based safety cases for AI scheming","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CR","authors_text":"Alexander Meinke, Bilal Chughtai, Buck Shlegeris, Charlotte Stix, Dan Braun, Daniel Kokotajlo, David Lindner, J\\'er\\'emy Scheurer, Joshua Clymer, Lucius Bushnaq, Marius Hobbhahn, Mikita Balesni, Nicholas Goldowsky-Dill, Owain Evans, Rusheb Shah, Tomek Korbak","submitted_at":"2024-10-29T17:55:29Z","abstract_excerpt":"We sketch how developers of frontier AI systems could construct a structured rationale -- a 'safety case' -- that an AI system is unlikely to cause catastrophic outcomes through scheming. Scheming is a potential threat model where AI systems could pursue misaligned goals covertly, hiding their true capabilities and objectives. In this report, we propose three arguments that safety cases could use in relation to scheming. For each argument we sketch how evidence could be gathered from empirical evaluations, and what assumptions would need to be met to provide strong assurance. First, developers"},"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":"2411.03336","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2024-10-29T17:55:29Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"d70d737ad7f4b7e5de8b24ad4d3be5ec360ad775de775a94f25ac699dbd78a32","abstract_canon_sha256":"8785c12312ee8a3b70aa23a5eda9bd5728e3b7ad5365bd7dd41c781db65827b8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:32:13.802350Z","signature_b64":"FIQIVCcGduks///UzbodR7tLaDuquHjoMC9zneDRg+lFe4gnlOrdCHh3Q3XFDg6hTLNJbZcCOJKsagDJUtYvBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3c99fa83ed82ea1cfbfa167c52104202ad164850c61ea158d7b6f61ef437e8af","last_reissued_at":"2026-07-05T09:32:13.800171Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:32:13.800171Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Towards evaluations-based safety cases for AI scheming","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CR","authors_text":"Alexander Meinke, Bilal Chughtai, Buck Shlegeris, Charlotte Stix, Dan Braun, Daniel Kokotajlo, David Lindner, J\\'er\\'emy Scheurer, Joshua Clymer, Lucius Bushnaq, Marius Hobbhahn, Mikita Balesni, Nicholas Goldowsky-Dill, Owain Evans, Rusheb Shah, Tomek Korbak","submitted_at":"2024-10-29T17:55:29Z","abstract_excerpt":"We sketch how developers of frontier AI systems could construct a structured rationale -- a 'safety case' -- that an AI system is unlikely to cause catastrophic outcomes through scheming. Scheming is a potential threat model where AI systems could pursue misaligned goals covertly, hiding their true capabilities and objectives. In this report, we propose three arguments that safety cases could use in relation to scheming. For each argument we sketch how evidence could be gathered from empirical evaluations, and what assumptions would need to be met to provide strong assurance. First, developers"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2411.03336","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/2411.03336/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":"2411.03336","created_at":"2026-07-05T09:32:13.801361+00:00"},{"alias_kind":"arxiv_version","alias_value":"2411.03336v2","created_at":"2026-07-05T09:32:13.801361+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2411.03336","created_at":"2026-07-05T09:32:13.801361+00:00"},{"alias_kind":"pith_short_12","alias_value":"HSM7VA7NQLVB","created_at":"2026-07-05T09:32:13.801361+00:00"},{"alias_kind":"pith_short_16","alias_value":"HSM7VA7NQLVBZ672","created_at":"2026-07-05T09:32:13.801361+00:00"},{"alias_kind":"pith_short_8","alias_value":"HSM7VA7N","created_at":"2026-07-05T09:32:13.801361+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2507.11473","citing_title":"Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2510.12826","citing_title":"Scheming Ability in LLM-to-LLM Strategic Interactions","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2512.01166","citing_title":"Evaluating AI Providers' Frontier Safety Frameworks","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2412.04984","citing_title":"Frontier Models are Capable of In-context Scheming","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2604.09104","citing_title":"Scheming in the wild: detecting real-world AI scheming incidents with open-source intelligence","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2604.13301","citing_title":"Honeypot Protocol","ref_index":1,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/HSM7VA7NQLVBZ672CZ6FEECCAK","json":"https://pith.science/pith/HSM7VA7NQLVBZ672CZ6FEECCAK.json","graph_json":"https://pith.science/api/pith-number/HSM7VA7NQLVBZ672CZ6FEECCAK/graph.json","events_json":"https://pith.science/api/pith-number/HSM7VA7NQLVBZ672CZ6FEECCAK/events.json","paper":"https://pith.science/paper/HSM7VA7N"},"agent_actions":{"view_html":"https://pith.science/pith/HSM7VA7NQLVBZ672CZ6FEECCAK","download_json":"https://pith.science/pith/HSM7VA7NQLVBZ672CZ6FEECCAK.json","view_paper":"https://pith.science/paper/HSM7VA7N","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2411.03336&json=true","fetch_graph":"https://pith.science/api/pith-number/HSM7VA7NQLVBZ672CZ6FEECCAK/graph.json","fetch_events":"https://pith.science/api/pith-number/HSM7VA7NQLVBZ672CZ6FEECCAK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/HSM7VA7NQLVBZ672CZ6FEECCAK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/HSM7VA7NQLVBZ672CZ6FEECCAK/action/storage_attestation","attest_author":"https://pith.science/pith/HSM7VA7NQLVBZ672CZ6FEECCAK/action/author_attestation","sign_citation":"https://pith.science/pith/HSM7VA7NQLVBZ672CZ6FEECCAK/action/citation_signature","submit_replication":"https://pith.science/pith/HSM7VA7NQLVBZ672CZ6FEECCAK/action/replication_record"}},"created_at":"2026-07-05T09:32:13.801361+00:00","updated_at":"2026-07-05T09:32:13.801361+00:00"}