{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:EAPFT73CSG2FZZQ3JQIURFRK53","short_pith_number":"pith:EAPFT73C","schema_version":"1.0","canonical_sha256":"201e59ff6291b45ce61b4c1148962aeeffa4050bf7764639aa77d317cd93e6b6","source":{"kind":"arxiv","id":"2302.05733","version":1},"attestation_state":"computed","paper":{"title":"Exploiting Programmatic Behavior of LLMs: Dual-Use Through Standard Security Attacks","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CR","authors_text":"Carlos Guestrin, Daniel Kang, Ion Stoica, Matei Zaharia, Tatsunori Hashimoto, Xuechen Li","submitted_at":"2023-02-11T15:57:44Z","abstract_excerpt":"Recent advances in instruction-following large language models (LLMs) have led to dramatic improvements in a range of NLP tasks. Unfortunately, we find that the same improved capabilities amplify the dual-use risks for malicious purposes of these models. Dual-use is difficult to prevent as instruction-following capabilities now enable standard attacks from computer security. The capabilities of these instruction-following LLMs provide strong economic incentives for dual-use by malicious actors. In particular, we show that instruction-following LLMs can produce targeted malicious content, inclu"},"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":"2302.05733","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2023-02-11T15:57:44Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"b2dd0a881ec75141bdc59ae20e468ea45432a7fbe2e5509adc433ae1dbf94c02","abstract_canon_sha256":"c4c70eb05df1f88b4e7b3f79fd8a832cfc6c803ef0631d2aeecf37e39b0bf791"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:40:57.693041Z","signature_b64":"jnwCfcZZwgXCfsQ2oJrXI5LfwigKWplMU4nRere2482IUBwkT3g+t8bxHvIy+e/dkgq8A/PHLjjSM4bSQyuXCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"201e59ff6291b45ce61b4c1148962aeeffa4050bf7764639aa77d317cd93e6b6","last_reissued_at":"2026-07-05T05:40:57.692606Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:40:57.692606Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Exploiting Programmatic Behavior of LLMs: Dual-Use Through Standard Security Attacks","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CR","authors_text":"Carlos Guestrin, Daniel Kang, Ion Stoica, Matei Zaharia, Tatsunori Hashimoto, Xuechen Li","submitted_at":"2023-02-11T15:57:44Z","abstract_excerpt":"Recent advances in instruction-following large language models (LLMs) have led to dramatic improvements in a range of NLP tasks. Unfortunately, we find that the same improved capabilities amplify the dual-use risks for malicious purposes of these models. Dual-use is difficult to prevent as instruction-following capabilities now enable standard attacks from computer security. The capabilities of these instruction-following LLMs provide strong economic incentives for dual-use by malicious actors. In particular, we show that instruction-following LLMs can produce targeted malicious content, inclu"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2302.05733","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/2302.05733/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":"2302.05733","created_at":"2026-07-05T05:40:57.692662+00:00"},{"alias_kind":"arxiv_version","alias_value":"2302.05733v1","created_at":"2026-07-05T05:40:57.692662+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2302.05733","created_at":"2026-07-05T05:40:57.692662+00:00"},{"alias_kind":"pith_short_12","alias_value":"EAPFT73CSG2F","created_at":"2026-07-05T05:40:57.692662+00:00"},{"alias_kind":"pith_short_16","alias_value":"EAPFT73CSG2FZZQ3","created_at":"2026-07-05T05:40:57.692662+00:00"},{"alias_kind":"pith_short_8","alias_value":"EAPFT73C","created_at":"2026-07-05T05:40:57.692662+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":16,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.25097","citing_title":"Speculative Decoding at Temperature Zero: A Scoped Safety-Invariance Screen with a 48,072-Sample Expansion","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2606.30783","citing_title":"Security--Fidelity Tradeoffs: The Hidden Cost of Prompt Injection Defense","ref_index":107,"is_internal_anchor":false},{"citing_arxiv_id":"2606.31639","citing_title":"A Lifecycle and Application-Stack Survey of Large Language Model Vulnerabilities: Attacks, Risks, Defenses, and Open Problems","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2402.06922","citing_title":"Whispers in the Machine: Confidentiality in Agentic Systems","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2410.15362","citing_title":"Faster-GCG: Efficient Discrete Optimization Jailbreak Attacks against Aligned Large Language Models","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2404.08144","citing_title":"LLM Agents can Autonomously Exploit One-day Vulnerabilities","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2308.03825","citing_title":"\"Do Anything Now\": Characterizing and Evaluating In-The-Wild Jailbreak Prompts on Large Language Models","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2310.06987","citing_title":"Catastrophic Jailbreak of Open-source LLMs via Exploiting Generation","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2402.10260","citing_title":"A StrongREJECT for Empty Jailbreaks","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2410.07283","citing_title":"Prompt Infection: LLM-to-LLM Prompt Injection within Multi-Agent Systems","ref_index":64,"is_internal_anchor":false},{"citing_arxiv_id":"2407.04295","citing_title":"Jailbreak Attacks and Defenses Against Large Language Models: A Survey","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2307.02483","citing_title":"Jailbroken: How Does LLM Safety Training Fail?","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2309.00614","citing_title":"Baseline Defenses for Adversarial Attacks Against Aligned Language Models","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2310.04451","citing_title":"AutoDAN: Generating Stealthy Jailbreak Prompts on Aligned Large Language Models","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2604.27001","citing_title":"An Empirical Security Evaluation of LLM-Generated Cryptographic Rust Code","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2605.00348","citing_title":"Block-wise Codeword Embedding for Reliable Multi-bit Text Watermarking","ref_index":58,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/EAPFT73CSG2FZZQ3JQIURFRK53","json":"https://pith.science/pith/EAPFT73CSG2FZZQ3JQIURFRK53.json","graph_json":"https://pith.science/api/pith-number/EAPFT73CSG2FZZQ3JQIURFRK53/graph.json","events_json":"https://pith.science/api/pith-number/EAPFT73CSG2FZZQ3JQIURFRK53/events.json","paper":"https://pith.science/paper/EAPFT73C"},"agent_actions":{"view_html":"https://pith.science/pith/EAPFT73CSG2FZZQ3JQIURFRK53","download_json":"https://pith.science/pith/EAPFT73CSG2FZZQ3JQIURFRK53.json","view_paper":"https://pith.science/paper/EAPFT73C","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2302.05733&json=true","fetch_graph":"https://pith.science/api/pith-number/EAPFT73CSG2FZZQ3JQIURFRK53/graph.json","fetch_events":"https://pith.science/api/pith-number/EAPFT73CSG2FZZQ3JQIURFRK53/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/EAPFT73CSG2FZZQ3JQIURFRK53/action/timestamp_anchor","attest_storage":"https://pith.science/pith/EAPFT73CSG2FZZQ3JQIURFRK53/action/storage_attestation","attest_author":"https://pith.science/pith/EAPFT73CSG2FZZQ3JQIURFRK53/action/author_attestation","sign_citation":"https://pith.science/pith/EAPFT73CSG2FZZQ3JQIURFRK53/action/citation_signature","submit_replication":"https://pith.science/pith/EAPFT73CSG2FZZQ3JQIURFRK53/action/replication_record"}},"created_at":"2026-07-05T05:40:57.692662+00:00","updated_at":"2026-07-05T05:40:57.692662+00:00"}