{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:ANYQJS5CBM724YZIRMIM5Y7IWJ","short_pith_number":"pith:ANYQJS5C","schema_version":"1.0","canonical_sha256":"037104cba20b3fae63288b10cee3e8b27d1a0f747e521a5bc044605019bb902e","source":{"kind":"arxiv","id":"2606.29657","version":1},"attestation_state":"computed","paper":{"title":"Safety from Honesty in a Disinterested AI Predictor","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.AI","authors_text":"Adam Oberman, Anna Gaven\\v{c}iak, Aton Kamanda, Damiano Fornasiere, David Hyland, Francis Rhys Ward, Gael Gendron, Iulian Serban, Jacob Livingston Slosser, Joumana Ghosn, Michael Cohen, Oliver Richardson, Rory Svarc, Tom\\'a\\v{s} Gaven\\v{c}iak, Vincent Mai, Yoshua Bengio","submitted_at":"2026-06-28T23:55:29Z","abstract_excerpt":"As AI systems become more capable, training procedures that optimize for downstream outcomes risk introducing implicit agency: goal-directed behavior that designers never specified. We present a formal safety argument for the Scientist AI (SAI) Predictor, trained to approximate the Bayesian posterior conditioned on a dataset of \"epistemically contextualized\" natural-language statements. We argue that such a Predictor can honestly predict agents, actions, and their consequences without itself being an agent that selects outputs to achieve goals. This rests on data representation and on the trai"},"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":"2606.29657","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-06-28T23:55:29Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"1c55f46a7c465d64dcf9060c266a2403aaadcd47f8d556a4c9a86f6f0597cb85","abstract_canon_sha256":"10a24b803ed74ef8958689f7e067ba4fecaf6347b4f149427bb365854a9b77b3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-30T02:17:29.201781Z","signature_b64":"zeSnVnuQ4MDq4EgrqCOYQvwuua6dmcRAmo7dF9g2r+nYlhGRldQ8cFpYb46vNY5w664VomeHsmhN8LVIxwMSAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"037104cba20b3fae63288b10cee3e8b27d1a0f747e521a5bc044605019bb902e","last_reissued_at":"2026-06-30T02:17:29.201261Z","signature_status":"signed_v1","first_computed_at":"2026-06-30T02:17:29.201261Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Safety from Honesty in a Disinterested AI Predictor","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.AI","authors_text":"Adam Oberman, Anna Gaven\\v{c}iak, Aton Kamanda, Damiano Fornasiere, David Hyland, Francis Rhys Ward, Gael Gendron, Iulian Serban, Jacob Livingston Slosser, Joumana Ghosn, Michael Cohen, Oliver Richardson, Rory Svarc, Tom\\'a\\v{s} Gaven\\v{c}iak, Vincent Mai, Yoshua Bengio","submitted_at":"2026-06-28T23:55:29Z","abstract_excerpt":"As AI systems become more capable, training procedures that optimize for downstream outcomes risk introducing implicit agency: goal-directed behavior that designers never specified. We present a formal safety argument for the Scientist AI (SAI) Predictor, trained to approximate the Bayesian posterior conditioned on a dataset of \"epistemically contextualized\" natural-language statements. We argue that such a Predictor can honestly predict agents, actions, and their consequences without itself being an agent that selects outputs to achieve goals. This rests on data representation and on the trai"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.29657","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/2606.29657/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":"2606.29657","created_at":"2026-06-30T02:17:29.201340+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.29657v1","created_at":"2026-06-30T02:17:29.201340+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.29657","created_at":"2026-06-30T02:17:29.201340+00:00"},{"alias_kind":"pith_short_12","alias_value":"ANYQJS5CBM72","created_at":"2026-06-30T02:17:29.201340+00:00"},{"alias_kind":"pith_short_16","alias_value":"ANYQJS5CBM724YZI","created_at":"2026-06-30T02:17:29.201340+00:00"},{"alias_kind":"pith_short_8","alias_value":"ANYQJS5C","created_at":"2026-06-30T02:17:29.201340+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ANYQJS5CBM724YZIRMIM5Y7IWJ","json":"https://pith.science/pith/ANYQJS5CBM724YZIRMIM5Y7IWJ.json","graph_json":"https://pith.science/api/pith-number/ANYQJS5CBM724YZIRMIM5Y7IWJ/graph.json","events_json":"https://pith.science/api/pith-number/ANYQJS5CBM724YZIRMIM5Y7IWJ/events.json","paper":"https://pith.science/paper/ANYQJS5C"},"agent_actions":{"view_html":"https://pith.science/pith/ANYQJS5CBM724YZIRMIM5Y7IWJ","download_json":"https://pith.science/pith/ANYQJS5CBM724YZIRMIM5Y7IWJ.json","view_paper":"https://pith.science/paper/ANYQJS5C","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.29657&json=true","fetch_graph":"https://pith.science/api/pith-number/ANYQJS5CBM724YZIRMIM5Y7IWJ/graph.json","fetch_events":"https://pith.science/api/pith-number/ANYQJS5CBM724YZIRMIM5Y7IWJ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ANYQJS5CBM724YZIRMIM5Y7IWJ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ANYQJS5CBM724YZIRMIM5Y7IWJ/action/storage_attestation","attest_author":"https://pith.science/pith/ANYQJS5CBM724YZIRMIM5Y7IWJ/action/author_attestation","sign_citation":"https://pith.science/pith/ANYQJS5CBM724YZIRMIM5Y7IWJ/action/citation_signature","submit_replication":"https://pith.science/pith/ANYQJS5CBM724YZIRMIM5Y7IWJ/action/replication_record"}},"created_at":"2026-06-30T02:17:29.201340+00:00","updated_at":"2026-06-30T02:17:29.201340+00:00"}