{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:AYI57RKG6PLI6FYD2DDFFVRYQ2","short_pith_number":"pith:AYI57RKG","schema_version":"1.0","canonical_sha256":"0611dfc546f3d68f1703d0c652d638869c032ca1927b9a326516086017499e1b","source":{"kind":"arxiv","id":"2606.05614","version":1},"attestation_state":"computed","paper":{"title":"Safety Paradox: How Enhanced Safety Awareness Leaves LLMs Vulnerable to Posterior Attack","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Hai V. Le, Long P. Hoang, Shaoyang Xu, Wei Lu, Wenxuan Zhang","submitted_at":"2026-06-04T02:36:41Z","abstract_excerpt":"Large language models (LLMs) are rigorously aligned to refuse harmful requests, a process that inherently cultivates a latent capacity to evaluate and recognize unsafe content. In this work, we reveal that this advanced safety awareness inadvertently introduces a fatal vulnerability. We introduce Posterior Attack, a single-query jailbreak that bypasses guardrails by prompting the model to generate the exact harmful response its internal classifier would normally flag as unsafe. Through extensive empirical evaluation across 30 open-source LLMs (up to 35B parameters in size) and frontier models "},"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":"2606.05614","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-06-04T02:36:41Z","cross_cats_sorted":[],"title_canon_sha256":"68c9cceef9d821975e3a1f951bc1884e1e15dee9384f3fcd1f50e0463799e660","abstract_canon_sha256":"bd69cfd2004eb5d7eadca2b180c06e8a4e37440c87e62e0358357439c57ad83a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-05T01:14:56.485232Z","signature_b64":"E/e0QGR7kzb5+I+ZZX/c5zYs8H9uJQ4/BFGXRxZf4OJODMPTwIrgRI+avVkFbL4fPXAMRR9VXyO2QJF+Kr+ABg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0611dfc546f3d68f1703d0c652d638869c032ca1927b9a326516086017499e1b","last_reissued_at":"2026-06-05T01:14:56.484785Z","signature_status":"signed_v1","first_computed_at":"2026-06-05T01:14:56.484785Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Safety Paradox: How Enhanced Safety Awareness Leaves LLMs Vulnerable to Posterior Attack","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Hai V. Le, Long P. Hoang, Shaoyang Xu, Wei Lu, Wenxuan Zhang","submitted_at":"2026-06-04T02:36:41Z","abstract_excerpt":"Large language models (LLMs) are rigorously aligned to refuse harmful requests, a process that inherently cultivates a latent capacity to evaluate and recognize unsafe content. In this work, we reveal that this advanced safety awareness inadvertently introduces a fatal vulnerability. We introduce Posterior Attack, a single-query jailbreak that bypasses guardrails by prompting the model to generate the exact harmful response its internal classifier would normally flag as unsafe. Through extensive empirical evaluation across 30 open-source LLMs (up to 35B parameters in size) and frontier models "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.05614","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/2606.05614/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":"2606.05614","created_at":"2026-06-05T01:14:56.484859+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.05614v1","created_at":"2026-06-05T01:14:56.484859+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.05614","created_at":"2026-06-05T01:14:56.484859+00:00"},{"alias_kind":"pith_short_12","alias_value":"AYI57RKG6PLI","created_at":"2026-06-05T01:14:56.484859+00:00"},{"alias_kind":"pith_short_16","alias_value":"AYI57RKG6PLI6FYD","created_at":"2026-06-05T01:14:56.484859+00:00"},{"alias_kind":"pith_short_8","alias_value":"AYI57RKG","created_at":"2026-06-05T01:14:56.484859+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/AYI57RKG6PLI6FYD2DDFFVRYQ2","json":"https://pith.science/pith/AYI57RKG6PLI6FYD2DDFFVRYQ2.json","graph_json":"https://pith.science/api/pith-number/AYI57RKG6PLI6FYD2DDFFVRYQ2/graph.json","events_json":"https://pith.science/api/pith-number/AYI57RKG6PLI6FYD2DDFFVRYQ2/events.json","paper":"https://pith.science/paper/AYI57RKG"},"agent_actions":{"view_html":"https://pith.science/pith/AYI57RKG6PLI6FYD2DDFFVRYQ2","download_json":"https://pith.science/pith/AYI57RKG6PLI6FYD2DDFFVRYQ2.json","view_paper":"https://pith.science/paper/AYI57RKG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.05614&json=true","fetch_graph":"https://pith.science/api/pith-number/AYI57RKG6PLI6FYD2DDFFVRYQ2/graph.json","fetch_events":"https://pith.science/api/pith-number/AYI57RKG6PLI6FYD2DDFFVRYQ2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/AYI57RKG6PLI6FYD2DDFFVRYQ2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/AYI57RKG6PLI6FYD2DDFFVRYQ2/action/storage_attestation","attest_author":"https://pith.science/pith/AYI57RKG6PLI6FYD2DDFFVRYQ2/action/author_attestation","sign_citation":"https://pith.science/pith/AYI57RKG6PLI6FYD2DDFFVRYQ2/action/citation_signature","submit_replication":"https://pith.science/pith/AYI57RKG6PLI6FYD2DDFFVRYQ2/action/replication_record"}},"created_at":"2026-06-05T01:14:56.484859+00:00","updated_at":"2026-06-05T01:14:56.484859+00:00"}