{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:XT6MUUMRIHXP5GNU4R4EXBAKPE","short_pith_number":"pith:XT6MUUMR","schema_version":"1.0","canonical_sha256":"bcfcca519141eefe99b4e4784b840a7915aed03336012fcb13bdd1ac8d257e2f","source":{"kind":"arxiv","id":"2505.22852","version":1},"attestation_state":"computed","paper":{"title":"Operationalizing CaMeL: Strengthening LLM Defenses for Enterprise Deployment","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CR","authors_text":"Emma Miller, Krti Tallam","submitted_at":"2025-05-28T20:35:24Z","abstract_excerpt":"CaMeL (Capabilities for Machine Learning) introduces a capability-based sandbox to mitigate prompt injection attacks in large language model (LLM) agents. While effective, CaMeL assumes a trusted user prompt, omits side-channel concerns, and incurs performance tradeoffs due to its dual-LLM design. This response identifies these issues and proposes engineering improvements to expand CaMeL's threat coverage and operational usability. We introduce: (1) prompt screening for initial inputs, (2) output auditing to detect instruction leakage, (3) a tiered-risk access model to balance usability and co"},"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":"2505.22852","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CR","submitted_at":"2025-05-28T20:35:24Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"ce476f78259db45f1f15c3893be24ec42ae3c97005cc7dbca22dcee08eb0e8eb","abstract_canon_sha256":"dcdc0b7e2bdad0b5f149651771158320797003216e92b7298630fafda304268d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:11:50.900796Z","signature_b64":"QDSxqjsUn+nhzEXpvfldJlcT2J58KhFwcbvLjEtsc9ifTMIvoHsDOy5pCsM3Rs7cImQdEcxOKx4omGZ2GNEMCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bcfcca519141eefe99b4e4784b840a7915aed03336012fcb13bdd1ac8d257e2f","last_reissued_at":"2026-07-05T11:11:50.900293Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:11:50.900293Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Operationalizing CaMeL: Strengthening LLM Defenses for Enterprise Deployment","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CR","authors_text":"Emma Miller, Krti Tallam","submitted_at":"2025-05-28T20:35:24Z","abstract_excerpt":"CaMeL (Capabilities for Machine Learning) introduces a capability-based sandbox to mitigate prompt injection attacks in large language model (LLM) agents. While effective, CaMeL assumes a trusted user prompt, omits side-channel concerns, and incurs performance tradeoffs due to its dual-LLM design. This response identifies these issues and proposes engineering improvements to expand CaMeL's threat coverage and operational usability. We introduce: (1) prompt screening for initial inputs, (2) output auditing to detect instruction leakage, (3) a tiered-risk access model to balance usability and co"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.22852","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/2505.22852/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":"2505.22852","created_at":"2026-07-05T11:11:50.900361+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.22852v1","created_at":"2026-07-05T11:11:50.900361+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.22852","created_at":"2026-07-05T11:11:50.900361+00:00"},{"alias_kind":"pith_short_12","alias_value":"XT6MUUMRIHXP","created_at":"2026-07-05T11:11:50.900361+00:00"},{"alias_kind":"pith_short_16","alias_value":"XT6MUUMRIHXP5GNU","created_at":"2026-07-05T11:11:50.900361+00:00"},{"alias_kind":"pith_short_8","alias_value":"XT6MUUMR","created_at":"2026-07-05T11:11:50.900361+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.12320","citing_title":"A Five-Plane Reference Architecture for Runtime Governance of Production AI Agents","ref_index":51,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XT6MUUMRIHXP5GNU4R4EXBAKPE","json":"https://pith.science/pith/XT6MUUMRIHXP5GNU4R4EXBAKPE.json","graph_json":"https://pith.science/api/pith-number/XT6MUUMRIHXP5GNU4R4EXBAKPE/graph.json","events_json":"https://pith.science/api/pith-number/XT6MUUMRIHXP5GNU4R4EXBAKPE/events.json","paper":"https://pith.science/paper/XT6MUUMR"},"agent_actions":{"view_html":"https://pith.science/pith/XT6MUUMRIHXP5GNU4R4EXBAKPE","download_json":"https://pith.science/pith/XT6MUUMRIHXP5GNU4R4EXBAKPE.json","view_paper":"https://pith.science/paper/XT6MUUMR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.22852&json=true","fetch_graph":"https://pith.science/api/pith-number/XT6MUUMRIHXP5GNU4R4EXBAKPE/graph.json","fetch_events":"https://pith.science/api/pith-number/XT6MUUMRIHXP5GNU4R4EXBAKPE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XT6MUUMRIHXP5GNU4R4EXBAKPE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XT6MUUMRIHXP5GNU4R4EXBAKPE/action/storage_attestation","attest_author":"https://pith.science/pith/XT6MUUMRIHXP5GNU4R4EXBAKPE/action/author_attestation","sign_citation":"https://pith.science/pith/XT6MUUMRIHXP5GNU4R4EXBAKPE/action/citation_signature","submit_replication":"https://pith.science/pith/XT6MUUMRIHXP5GNU4R4EXBAKPE/action/replication_record"}},"created_at":"2026-07-05T11:11:50.900361+00:00","updated_at":"2026-07-05T11:11:50.900361+00:00"}