{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:ULBMKJPA5HTHQTKY7KXOLZDILN","short_pith_number":"pith:ULBMKJPA","schema_version":"1.0","canonical_sha256":"a2c2c525e0e9e6784d58faaee5e4685b53a555c543e72b7c2aac025938a419f9","source":{"kind":"arxiv","id":"2606.28739","version":1},"attestation_state":"computed","paper":{"title":"Agent Safety Is Action Alignment","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Shawn Li, Yue Zhao","submitted_at":"2026-06-27T05:26:43Z","abstract_excerpt":"Large language models increasingly act as agents: they call tools, move money, delete records, and send messages on a user's behalf. To keep them safe, practitioners imported the chatbot-era recipe (train the model to refuse unsafe inputs) into the agentic setting, and treat the resulting capability loss as a manageable ``alignment tax.'' We argue this is a \\emph{category error}. Refusal is a primitive for \\emph{content safety}, where the harm is in the model's output and is therefore a learnable function of it. Agentic harm is different in kind: it lies not in any output but in the relation b"},"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":"2606.28739","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-06-27T05:26:43Z","cross_cats_sorted":[],"title_canon_sha256":"752279e7991b28a1bb0a11ea9e49017fd3fa56e4c3d898e41418315820754700","abstract_canon_sha256":"5c07351daacaca098554e800537a3e1159bd32a53a45aa5ff5dac875c02062ef"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-30T01:16:49.900475Z","signature_b64":"/NmUjNjl11XEawkxyIeMObvQ7GbT7Myv0mYiTzR2HQP8XMvDPCLzidNWoYo8gLonweu1+lOgA0KRyrDmxF0zDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a2c2c525e0e9e6784d58faaee5e4685b53a555c543e72b7c2aac025938a419f9","last_reissued_at":"2026-06-30T01:16:49.899962Z","signature_status":"signed_v1","first_computed_at":"2026-06-30T01:16:49.899962Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Agent Safety Is Action Alignment","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Shawn Li, Yue Zhao","submitted_at":"2026-06-27T05:26:43Z","abstract_excerpt":"Large language models increasingly act as agents: they call tools, move money, delete records, and send messages on a user's behalf. To keep them safe, practitioners imported the chatbot-era recipe (train the model to refuse unsafe inputs) into the agentic setting, and treat the resulting capability loss as a manageable ``alignment tax.'' We argue this is a \\emph{category error}. Refusal is a primitive for \\emph{content safety}, where the harm is in the model's output and is therefore a learnable function of it. Agentic harm is different in kind: it lies not in any output but in the relation b"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.28739","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/2606.28739/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":"2606.28739","created_at":"2026-06-30T01:16:49.900039+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.28739v1","created_at":"2026-06-30T01:16:49.900039+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.28739","created_at":"2026-06-30T01:16:49.900039+00:00"},{"alias_kind":"pith_short_12","alias_value":"ULBMKJPA5HTH","created_at":"2026-06-30T01:16:49.900039+00:00"},{"alias_kind":"pith_short_16","alias_value":"ULBMKJPA5HTHQTKY","created_at":"2026-06-30T01:16:49.900039+00:00"},{"alias_kind":"pith_short_8","alias_value":"ULBMKJPA","created_at":"2026-06-30T01:16:49.900039+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/ULBMKJPA5HTHQTKY7KXOLZDILN","json":"https://pith.science/pith/ULBMKJPA5HTHQTKY7KXOLZDILN.json","graph_json":"https://pith.science/api/pith-number/ULBMKJPA5HTHQTKY7KXOLZDILN/graph.json","events_json":"https://pith.science/api/pith-number/ULBMKJPA5HTHQTKY7KXOLZDILN/events.json","paper":"https://pith.science/paper/ULBMKJPA"},"agent_actions":{"view_html":"https://pith.science/pith/ULBMKJPA5HTHQTKY7KXOLZDILN","download_json":"https://pith.science/pith/ULBMKJPA5HTHQTKY7KXOLZDILN.json","view_paper":"https://pith.science/paper/ULBMKJPA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.28739&json=true","fetch_graph":"https://pith.science/api/pith-number/ULBMKJPA5HTHQTKY7KXOLZDILN/graph.json","fetch_events":"https://pith.science/api/pith-number/ULBMKJPA5HTHQTKY7KXOLZDILN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ULBMKJPA5HTHQTKY7KXOLZDILN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ULBMKJPA5HTHQTKY7KXOLZDILN/action/storage_attestation","attest_author":"https://pith.science/pith/ULBMKJPA5HTHQTKY7KXOLZDILN/action/author_attestation","sign_citation":"https://pith.science/pith/ULBMKJPA5HTHQTKY7KXOLZDILN/action/citation_signature","submit_replication":"https://pith.science/pith/ULBMKJPA5HTHQTKY7KXOLZDILN/action/replication_record"}},"created_at":"2026-06-30T01:16:49.900039+00:00","updated_at":"2026-06-30T01:16:49.900039+00:00"}