{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:MD7PLAZUSMPNLTU5BDUC7OLIBR","short_pith_number":"pith:MD7PLAZU","schema_version":"1.0","canonical_sha256":"60fef58334931ed5ce9d08e82fb9680c66328f07a11f41268ed2ef5bb967b4b6","source":{"kind":"arxiv","id":"2307.16382","version":3},"attestation_state":"computed","paper":{"title":"Does fine-tuning GPT-3 with the OpenAI API leak personally-identifiable information?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Albert Yu Sun, Arushi Saxena, Christian Lau, Eliott Zemour, Eric Lin, Udith Vaidyanathan, Vaikkunth Mugunthan","submitted_at":"2023-07-31T03:17:51Z","abstract_excerpt":"Machine learning practitioners often fine-tune generative pre-trained models like GPT-3 to improve model performance at specific tasks. Previous works, however, suggest that fine-tuned machine learning models memorize and emit sensitive information from the original fine-tuning dataset. Companies such as OpenAI offer fine-tuning services for their models, but no prior work has conducted a memorization attack on any closed-source models. In this work, we simulate a privacy attack on GPT-3 using OpenAI's fine-tuning API. Our objective is to determine if personally identifiable information (PII) "},"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":"2307.16382","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-07-31T03:17:51Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"506eb6b495b19c8d53f23750143e27839a9e5983d5595801f2398d7d47abc986","abstract_canon_sha256":"b0d431a1f7e393da236c5c3adb27b474e15f9d60ec487c9082d1ec56047dc125"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:08:17.761046Z","signature_b64":"/Z5/1Wel+Kg1dHvudR0ZGRoQiFCw+78HiqQPg4AizzK/gkUkvGIlv91qfx593UoJVM0hpbhlaEFnkXgVeFBIAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"60fef58334931ed5ce9d08e82fb9680c66328f07a11f41268ed2ef5bb967b4b6","last_reissued_at":"2026-07-05T08:08:17.760513Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:08:17.760513Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Does fine-tuning GPT-3 with the OpenAI API leak personally-identifiable information?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Albert Yu Sun, Arushi Saxena, Christian Lau, Eliott Zemour, Eric Lin, Udith Vaidyanathan, Vaikkunth Mugunthan","submitted_at":"2023-07-31T03:17:51Z","abstract_excerpt":"Machine learning practitioners often fine-tune generative pre-trained models like GPT-3 to improve model performance at specific tasks. Previous works, however, suggest that fine-tuned machine learning models memorize and emit sensitive information from the original fine-tuning dataset. Companies such as OpenAI offer fine-tuning services for their models, but no prior work has conducted a memorization attack on any closed-source models. In this work, we simulate a privacy attack on GPT-3 using OpenAI's fine-tuning API. Our objective is to determine if personally identifiable information (PII) "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2307.16382","kind":"arxiv","version":3},"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/2307.16382/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":"2307.16382","created_at":"2026-07-05T08:08:17.760570+00:00"},{"alias_kind":"arxiv_version","alias_value":"2307.16382v3","created_at":"2026-07-05T08:08:17.760570+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2307.16382","created_at":"2026-07-05T08:08:17.760570+00:00"},{"alias_kind":"pith_short_12","alias_value":"MD7PLAZUSMPN","created_at":"2026-07-05T08:08:17.760570+00:00"},{"alias_kind":"pith_short_16","alias_value":"MD7PLAZUSMPNLTU5","created_at":"2026-07-05T08:08:17.760570+00:00"},{"alias_kind":"pith_short_8","alias_value":"MD7PLAZU","created_at":"2026-07-05T08:08:17.760570+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.15425","citing_title":"Runtime-Structured Task Decomposition for Agentic Coding Systems","ref_index":37,"is_internal_anchor":false},{"citing_arxiv_id":"2604.24902","citing_title":"Safety Drift After Fine-Tuning: Evidence from High-Stakes Domains","ref_index":55,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/MD7PLAZUSMPNLTU5BDUC7OLIBR","json":"https://pith.science/pith/MD7PLAZUSMPNLTU5BDUC7OLIBR.json","graph_json":"https://pith.science/api/pith-number/MD7PLAZUSMPNLTU5BDUC7OLIBR/graph.json","events_json":"https://pith.science/api/pith-number/MD7PLAZUSMPNLTU5BDUC7OLIBR/events.json","paper":"https://pith.science/paper/MD7PLAZU"},"agent_actions":{"view_html":"https://pith.science/pith/MD7PLAZUSMPNLTU5BDUC7OLIBR","download_json":"https://pith.science/pith/MD7PLAZUSMPNLTU5BDUC7OLIBR.json","view_paper":"https://pith.science/paper/MD7PLAZU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2307.16382&json=true","fetch_graph":"https://pith.science/api/pith-number/MD7PLAZUSMPNLTU5BDUC7OLIBR/graph.json","fetch_events":"https://pith.science/api/pith-number/MD7PLAZUSMPNLTU5BDUC7OLIBR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MD7PLAZUSMPNLTU5BDUC7OLIBR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MD7PLAZUSMPNLTU5BDUC7OLIBR/action/storage_attestation","attest_author":"https://pith.science/pith/MD7PLAZUSMPNLTU5BDUC7OLIBR/action/author_attestation","sign_citation":"https://pith.science/pith/MD7PLAZUSMPNLTU5BDUC7OLIBR/action/citation_signature","submit_replication":"https://pith.science/pith/MD7PLAZUSMPNLTU5BDUC7OLIBR/action/replication_record"}},"created_at":"2026-07-05T08:08:17.760570+00:00","updated_at":"2026-07-05T08:08:17.760570+00:00"}