{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:64ZWKEDS4VIUSNDTPANX4GZ52Y","short_pith_number":"pith:64ZWKEDS","schema_version":"1.0","canonical_sha256":"f733651072e551493473781b7e1b3dd601855f714a4affefc680e58bc40291c3","source":{"kind":"arxiv","id":"2401.06373","version":2},"attestation_state":"computed","paper":{"title":"How Johnny Can Persuade LLMs to Jailbreak Them: Rethinking Persuasion to Challenge AI Safety by Humanizing LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Diyi Yang, Hongpeng Lin, Jingwen Zhang, Ruoxi Jia, Weiyan Shi, Yi Zeng","submitted_at":"2024-01-12T16:13:24Z","abstract_excerpt":"Most traditional AI safety research has approached AI models as machines and centered on algorithm-focused attacks developed by security experts. As large language models (LLMs) become increasingly common and competent, non-expert users can also impose risks during daily interactions. This paper introduces a new perspective to jailbreak LLMs as human-like communicators, to explore this overlooked intersection between everyday language interaction and AI safety. Specifically, we study how to persuade LLMs to jailbreak them. First, we propose a persuasion taxonomy derived from decades of social "},"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":"2401.06373","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-01-12T16:13:24Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"e95d45ac08d7e5fb69a43e029147f02c6d2c93a5a644823619bb14ad52459645","abstract_canon_sha256":"e6dd431c296cc89dde9fc3b6999d91d6d6779294929ba66d72899b7baf0fdb7e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:36:57.296209Z","signature_b64":"imCD/VC6sjHP1Qmhl5E5Bm9+TKYCz72boanFOEjrIC1sLGSgDlONPvHy1bPU1DTmwStJSZrr8/pn/vGj97KOBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f733651072e551493473781b7e1b3dd601855f714a4affefc680e58bc40291c3","last_reissued_at":"2026-07-05T07:36:57.295780Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:36:57.295780Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"How Johnny Can Persuade LLMs to Jailbreak Them: Rethinking Persuasion to Challenge AI Safety by Humanizing LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Diyi Yang, Hongpeng Lin, Jingwen Zhang, Ruoxi Jia, Weiyan Shi, Yi Zeng","submitted_at":"2024-01-12T16:13:24Z","abstract_excerpt":"Most traditional AI safety research has approached AI models as machines and centered on algorithm-focused attacks developed by security experts. As large language models (LLMs) become increasingly common and competent, non-expert users can also impose risks during daily interactions. This paper introduces a new perspective to jailbreak LLMs as human-like communicators, to explore this overlooked intersection between everyday language interaction and AI safety. Specifically, we study how to persuade LLMs to jailbreak them. First, we propose a persuasion taxonomy derived from decades of social "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2401.06373","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/2401.06373/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":"2401.06373","created_at":"2026-07-05T07:36:57.295835+00:00"},{"alias_kind":"arxiv_version","alias_value":"2401.06373v2","created_at":"2026-07-05T07:36:57.295835+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2401.06373","created_at":"2026-07-05T07:36:57.295835+00:00"},{"alias_kind":"pith_short_12","alias_value":"64ZWKEDS4VIU","created_at":"2026-07-05T07:36:57.295835+00:00"},{"alias_kind":"pith_short_16","alias_value":"64ZWKEDS4VIUSNDT","created_at":"2026-07-05T07:36:57.295835+00:00"},{"alias_kind":"pith_short_8","alias_value":"64ZWKEDS","created_at":"2026-07-05T07:36:57.295835+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":16,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.18193","citing_title":"A Red-Team Study of Anthropic Fable 5 & Opus 4.8 Models","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2606.05523","citing_title":"CHASE: Adversarial Red-Blue Teaming for Improving LLM Safety using Reinforcement Learning","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2606.00651","citing_title":"MESA: Improving MoE Safety Alignment via Decentralized Expertise","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2505.10846","citing_title":"AutoRAN: Automated Hijacking of Safety Reasoning in Large Reasoning Models","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2508.04204","citing_title":"ReasoningGuard: Safeguarding Large Reasoning Models with Inference-time Safety Aha Moments","ref_index":63,"is_internal_anchor":false},{"citing_arxiv_id":"2508.10880","citing_title":"Searching for Privacy Risks in LLM Agents via Simulation","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2511.02356","citing_title":"ASTRA: An Automated Framework for Strategy Discovery, Retrieval, and Evolution for Jailbreaking LLMs","ref_index":58,"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":92,"is_internal_anchor":false},{"citing_arxiv_id":"2402.10260","citing_title":"A StrongREJECT for Empty Jailbreaks","ref_index":41,"is_internal_anchor":false},{"citing_arxiv_id":"2404.01318","citing_title":"JailbreakBench: An Open Robustness Benchmark for Jailbreaking Large Language Models","ref_index":58,"is_internal_anchor":false},{"citing_arxiv_id":"2407.04295","citing_title":"Jailbreak Attacks and Defenses Against Large Language Models: A Survey","ref_index":109,"is_internal_anchor":false},{"citing_arxiv_id":"2604.04060","citing_title":"CoopGuard: Stateful Cooperative Agents Safeguarding LLMs Against Evolving Multi-Round Attacks","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2605.00267","citing_title":"Jailbroken Frontier Models Retain Their Capabilities","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2604.10403","citing_title":"Latent Instruction Representation Alignment: defending against jailbreaks, backdoors and undesired knowledge in LLMs","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2604.15780","citing_title":"Pruning Unsafe Tickets: A Resource-Efficient Framework for Safer and More Robust LLMs","ref_index":49,"is_internal_anchor":false},{"citing_arxiv_id":"2605.02647","citing_title":"ContextualJailbreak: Evolutionary Red-Teaming via Simulated Conversational Priming","ref_index":38,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/64ZWKEDS4VIUSNDTPANX4GZ52Y","json":"https://pith.science/pith/64ZWKEDS4VIUSNDTPANX4GZ52Y.json","graph_json":"https://pith.science/api/pith-number/64ZWKEDS4VIUSNDTPANX4GZ52Y/graph.json","events_json":"https://pith.science/api/pith-number/64ZWKEDS4VIUSNDTPANX4GZ52Y/events.json","paper":"https://pith.science/paper/64ZWKEDS"},"agent_actions":{"view_html":"https://pith.science/pith/64ZWKEDS4VIUSNDTPANX4GZ52Y","download_json":"https://pith.science/pith/64ZWKEDS4VIUSNDTPANX4GZ52Y.json","view_paper":"https://pith.science/paper/64ZWKEDS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2401.06373&json=true","fetch_graph":"https://pith.science/api/pith-number/64ZWKEDS4VIUSNDTPANX4GZ52Y/graph.json","fetch_events":"https://pith.science/api/pith-number/64ZWKEDS4VIUSNDTPANX4GZ52Y/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/64ZWKEDS4VIUSNDTPANX4GZ52Y/action/timestamp_anchor","attest_storage":"https://pith.science/pith/64ZWKEDS4VIUSNDTPANX4GZ52Y/action/storage_attestation","attest_author":"https://pith.science/pith/64ZWKEDS4VIUSNDTPANX4GZ52Y/action/author_attestation","sign_citation":"https://pith.science/pith/64ZWKEDS4VIUSNDTPANX4GZ52Y/action/citation_signature","submit_replication":"https://pith.science/pith/64ZWKEDS4VIUSNDTPANX4GZ52Y/action/replication_record"}},"created_at":"2026-07-05T07:36:57.295835+00:00","updated_at":"2026-07-05T07:36:57.295835+00:00"}