{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:PQG2PCLRYZU4DLGJSAH7C3ZQB5","short_pith_number":"pith:PQG2PCLR","schema_version":"1.0","canonical_sha256":"7c0da78971c669c1acc9900ff16f300f51a57aa42b42ac8c7ddb3b0966fc43d9","source":{"kind":"arxiv","id":"2311.04378","version":5},"attestation_state":"computed","paper":{"title":"Watermarks in the Sand: Impossibility of Strong Watermarking for Generative Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.CR"],"primary_cat":"cs.LG","authors_text":"Benjamin L. Edelman, Boaz Barak, Daniele Venturi, Danilo Francati, Giuseppe Ateniese, Hanlin Zhang","submitted_at":"2023-11-07T22:52:54Z","abstract_excerpt":"Watermarking generative models consists of planting a statistical signal (watermark) in a model's output so that it can be later verified that the output was generated by the given model. A strong watermarking scheme satisfies the property that a computationally bounded attacker cannot erase the watermark without causing significant quality degradation. In this paper, we study the (im)possibility of strong watermarking schemes. We prove that, under well-specified and natural assumptions, strong watermarking is impossible to achieve. This holds even in the private detection algorithm setting, w"},"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":"2311.04378","kind":"arxiv","version":5},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-11-07T22:52:54Z","cross_cats_sorted":["cs.CL","cs.CR"],"title_canon_sha256":"38586fe5887951ca880f403f06ccb2355090f62ac872c936c89f2225d6e66d8f","abstract_canon_sha256":"c4af5903284d1ce58bc7e2ac207efe55d988cd028896341f7c9e79417c4bba02"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:10:44.539427Z","signature_b64":"mQqUiG2lMXTNDCRQjyN3U+PRu4LT+xsJ/6yf3rLbeBsiKeAYsAIYqOz+ybAhMA+tfISBKhxhW4aSXI8nkDtJDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7c0da78971c669c1acc9900ff16f300f51a57aa42b42ac8c7ddb3b0966fc43d9","last_reissued_at":"2026-07-05T11:10:44.538958Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:10:44.538958Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Watermarks in the Sand: Impossibility of Strong Watermarking for Generative Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.CR"],"primary_cat":"cs.LG","authors_text":"Benjamin L. Edelman, Boaz Barak, Daniele Venturi, Danilo Francati, Giuseppe Ateniese, Hanlin Zhang","submitted_at":"2023-11-07T22:52:54Z","abstract_excerpt":"Watermarking generative models consists of planting a statistical signal (watermark) in a model's output so that it can be later verified that the output was generated by the given model. A strong watermarking scheme satisfies the property that a computationally bounded attacker cannot erase the watermark without causing significant quality degradation. In this paper, we study the (im)possibility of strong watermarking schemes. We prove that, under well-specified and natural assumptions, strong watermarking is impossible to achieve. This holds even in the private detection algorithm setting, w"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2311.04378","kind":"arxiv","version":5},"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/2311.04378/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":"2311.04378","created_at":"2026-07-05T11:10:44.539013+00:00"},{"alias_kind":"arxiv_version","alias_value":"2311.04378v5","created_at":"2026-07-05T11:10:44.539013+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2311.04378","created_at":"2026-07-05T11:10:44.539013+00:00"},{"alias_kind":"pith_short_12","alias_value":"PQG2PCLRYZU4","created_at":"2026-07-05T11:10:44.539013+00:00"},{"alias_kind":"pith_short_16","alias_value":"PQG2PCLRYZU4DLGJ","created_at":"2026-07-05T11:10:44.539013+00:00"},{"alias_kind":"pith_short_8","alias_value":"PQG2PCLR","created_at":"2026-07-05T11:10:44.539013+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":9,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.00621","citing_title":"Authenticity Debt and the Synthetic Content Threat Landscape: A Layered Framework for Trust, Provenance, and IP Governance in the Generative AI Era","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2303.11156","citing_title":"Can AI-Generated Text be Reliably Detected?","ref_index":56,"is_internal_anchor":false},{"citing_arxiv_id":"2605.16471","citing_title":"From AI-Generated Content to Agentic Action: Security and Safety Threats in Generative AI","ref_index":154,"is_internal_anchor":false},{"citing_arxiv_id":"2605.16035","citing_title":"Who Owns This Agent? Tracing AI Agents Back to Their Owners","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2509.10577","citing_title":"The Coding Limits of Robust Watermarking for Generative Models","ref_index":32,"is_internal_anchor":false},{"citing_arxiv_id":"2510.18333","citing_title":"Position: LLM Watermarking Should Align Stakeholders' Incentives for Practical Adoption","ref_index":67,"is_internal_anchor":false},{"citing_arxiv_id":"2512.10100","citing_title":"Robust AI Security and Alignment: A Sisyphean Endeavor?","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2604.02343","citing_title":"Haiku to Opus in Just 10 bits: LLMs Unlock Massive Compression Gains","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05443","citing_title":"SLAM: Structural Linguistic Activation Marking for Language Models","ref_index":27,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PQG2PCLRYZU4DLGJSAH7C3ZQB5","json":"https://pith.science/pith/PQG2PCLRYZU4DLGJSAH7C3ZQB5.json","graph_json":"https://pith.science/api/pith-number/PQG2PCLRYZU4DLGJSAH7C3ZQB5/graph.json","events_json":"https://pith.science/api/pith-number/PQG2PCLRYZU4DLGJSAH7C3ZQB5/events.json","paper":"https://pith.science/paper/PQG2PCLR"},"agent_actions":{"view_html":"https://pith.science/pith/PQG2PCLRYZU4DLGJSAH7C3ZQB5","download_json":"https://pith.science/pith/PQG2PCLRYZU4DLGJSAH7C3ZQB5.json","view_paper":"https://pith.science/paper/PQG2PCLR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2311.04378&json=true","fetch_graph":"https://pith.science/api/pith-number/PQG2PCLRYZU4DLGJSAH7C3ZQB5/graph.json","fetch_events":"https://pith.science/api/pith-number/PQG2PCLRYZU4DLGJSAH7C3ZQB5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PQG2PCLRYZU4DLGJSAH7C3ZQB5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PQG2PCLRYZU4DLGJSAH7C3ZQB5/action/storage_attestation","attest_author":"https://pith.science/pith/PQG2PCLRYZU4DLGJSAH7C3ZQB5/action/author_attestation","sign_citation":"https://pith.science/pith/PQG2PCLRYZU4DLGJSAH7C3ZQB5/action/citation_signature","submit_replication":"https://pith.science/pith/PQG2PCLRYZU4DLGJSAH7C3ZQB5/action/replication_record"}},"created_at":"2026-07-05T11:10:44.539013+00:00","updated_at":"2026-07-05T11:10:44.539013+00:00"}