{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:REQ33IVCKJ4CLU5EGMAKSUWWY3","short_pith_number":"pith:REQ33IVC","schema_version":"1.0","canonical_sha256":"8921bda2a2527825d3a43300a952d6c6d2f73ed977583b74e2893699cfca7d6d","source":{"kind":"arxiv","id":"2310.17884","version":2},"attestation_state":"computed","paper":{"title":"Can LLMs Keep a Secret? Testing Privacy Implications of Language Models via Contextual Integrity Theory","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.CR"],"primary_cat":"cs.AI","authors_text":"Hyunwoo Kim, Maarten Sap, Niloofar Mireshghallah, Reza Shokri, Xuhui Zhou, Yejin Choi, Yulia Tsvetkov","submitted_at":"2023-10-27T04:15:30Z","abstract_excerpt":"The interactive use of large language models (LLMs) in AI assistants (at work, home, etc.) introduces a new set of inference-time privacy risks: LLMs are fed different types of information from multiple sources in their inputs and are expected to reason about what to share in their outputs, for what purpose and with whom, within a given context. In this work, we draw attention to the highly critical yet overlooked notion of contextual privacy by proposing ConfAIde, a benchmark designed to identify critical weaknesses in the privacy reasoning capabilities of instruction-tuned LLMs. Our experime"},"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":"2310.17884","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2023-10-27T04:15:30Z","cross_cats_sorted":["cs.CL","cs.CR"],"title_canon_sha256":"b8e934f7e6e2b2d51359ea47ddee42be359ee7dd8f71be0f0e023a432a121834","abstract_canon_sha256":"1811ea3d940f7cdc19100737a0ff19eca32ad01eee9f74221e014c37af5a9284"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:38:03.820374Z","signature_b64":"81NyoClRQTafrIR81KmtzdSz5fS9i9HZWO9i0qqqmPEPURBU0mv4FGVslSRjVyDFRrNjATMVHv/pV7AipEQ2BQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8921bda2a2527825d3a43300a952d6c6d2f73ed977583b74e2893699cfca7d6d","last_reissued_at":"2026-07-05T08:38:03.819762Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:38:03.819762Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Can LLMs Keep a Secret? Testing Privacy Implications of Language Models via Contextual Integrity Theory","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.CR"],"primary_cat":"cs.AI","authors_text":"Hyunwoo Kim, Maarten Sap, Niloofar Mireshghallah, Reza Shokri, Xuhui Zhou, Yejin Choi, Yulia Tsvetkov","submitted_at":"2023-10-27T04:15:30Z","abstract_excerpt":"The interactive use of large language models (LLMs) in AI assistants (at work, home, etc.) introduces a new set of inference-time privacy risks: LLMs are fed different types of information from multiple sources in their inputs and are expected to reason about what to share in their outputs, for what purpose and with whom, within a given context. In this work, we draw attention to the highly critical yet overlooked notion of contextual privacy by proposing ConfAIde, a benchmark designed to identify critical weaknesses in the privacy reasoning capabilities of instruction-tuned LLMs. Our experime"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2310.17884","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/2310.17884/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":"2310.17884","created_at":"2026-07-05T08:38:03.819830+00:00"},{"alias_kind":"arxiv_version","alias_value":"2310.17884v2","created_at":"2026-07-05T08:38:03.819830+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2310.17884","created_at":"2026-07-05T08:38:03.819830+00:00"},{"alias_kind":"pith_short_12","alias_value":"REQ33IVCKJ4C","created_at":"2026-07-05T08:38:03.819830+00:00"},{"alias_kind":"pith_short_16","alias_value":"REQ33IVCKJ4CLU5E","created_at":"2026-07-05T08:38:03.819830+00:00"},{"alias_kind":"pith_short_8","alias_value":"REQ33IVC","created_at":"2026-07-05T08:38:03.819830+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":16,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.06371","citing_title":"The Impact of Security and Privacy Controls on Users' Emotional Engagement with Generative AI Chatbots","ref_index":52,"is_internal_anchor":true},{"citing_arxiv_id":"2606.24535","citing_title":"Governed Shared Memory for Multi-Agent LLM Systems","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2606.26627","citing_title":"Agents That Know Too Much: A Data-Centric Survey of Privacy in LLM Agents","ref_index":81,"is_internal_anchor":false},{"citing_arxiv_id":"2606.12341","citing_title":"OCELOT: Inference-Leakage Budgets for Privacy-Preserving LLM Agents","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2606.07175","citing_title":"Seeing Without Exposing: Adaptive Privacy Control for Open-World, Context-Hungry MLLMs","ref_index":35,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09823","citing_title":"CalBench: Evaluating Coordination-Privacy Trade-offs in Multi-Agent LLMs","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2606.30383","citing_title":"Whose Side Is Your Agent On? Multi-Party Principal Loyalty in LLM Agents","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2605.16630","citing_title":"PrivScope: Task-scoped Disclosure Control for Hybrid Agentic Systems","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17830","citing_title":"Remembering More, Risking More: Longitudinal Safety Risks in Memory-Equipped LLM Agents","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2402.17177","citing_title":"Sora: A Review on Background, Technology, Limitations, and Opportunities of Large Vision Models","ref_index":132,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10013","citing_title":"When Are LLM Inferences Acceptable? User Reactions and Control Preferences for Inferred Personal Information","ref_index":39,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10794","citing_title":"Can You Keep a Secret? Involuntary Information Leakage in Language Model Writing","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08647","citing_title":"AgentCollabBench: Diagnosing When Good Agents Make Bad Collaborators","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05340","citing_title":"How Far Are VLMs from Privacy Awareness in the Physical World? An Empirical Study","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05340","citing_title":"How Far Are VLMs from Privacy Awareness in the Physical World? An Empirical Study","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2604.16521","citing_title":"CAMP: Cumulative Agentic Masking and Pruning for Privacy Protection in Multi-Turn LLM Conversations","ref_index":20,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/REQ33IVCKJ4CLU5EGMAKSUWWY3","json":"https://pith.science/pith/REQ33IVCKJ4CLU5EGMAKSUWWY3.json","graph_json":"https://pith.science/api/pith-number/REQ33IVCKJ4CLU5EGMAKSUWWY3/graph.json","events_json":"https://pith.science/api/pith-number/REQ33IVCKJ4CLU5EGMAKSUWWY3/events.json","paper":"https://pith.science/paper/REQ33IVC"},"agent_actions":{"view_html":"https://pith.science/pith/REQ33IVCKJ4CLU5EGMAKSUWWY3","download_json":"https://pith.science/pith/REQ33IVCKJ4CLU5EGMAKSUWWY3.json","view_paper":"https://pith.science/paper/REQ33IVC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2310.17884&json=true","fetch_graph":"https://pith.science/api/pith-number/REQ33IVCKJ4CLU5EGMAKSUWWY3/graph.json","fetch_events":"https://pith.science/api/pith-number/REQ33IVCKJ4CLU5EGMAKSUWWY3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/REQ33IVCKJ4CLU5EGMAKSUWWY3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/REQ33IVCKJ4CLU5EGMAKSUWWY3/action/storage_attestation","attest_author":"https://pith.science/pith/REQ33IVCKJ4CLU5EGMAKSUWWY3/action/author_attestation","sign_citation":"https://pith.science/pith/REQ33IVCKJ4CLU5EGMAKSUWWY3/action/citation_signature","submit_replication":"https://pith.science/pith/REQ33IVCKJ4CLU5EGMAKSUWWY3/action/replication_record"}},"created_at":"2026-07-05T08:38:03.819830+00:00","updated_at":"2026-07-05T08:38:03.819830+00:00"}