{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2022:3RI5ZPJT67FOOWE5DHSCSYQKSK","short_pith_number":"pith:3RI5ZPJT","canonical_record":{"source":{"id":"2212.11415","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2022-12-21T23:52:42Z","cross_cats_sorted":[],"title_canon_sha256":"cb357e916b270b668e8118ba71ca10d6ecbe55c66d6f38341ce06eca487a0869","abstract_canon_sha256":"fcd1d203779ff3901bd8c7b74f2254203ed30a98b325c2c1f4e144f24a8ca2a1"},"schema_version":"1.0"},"canonical_sha256":"dc51dcbd33f7cae7589d19e429620a929a0b5a07cd70296a3cb38887c361c050","source":{"kind":"arxiv","id":"2212.11415","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2212.11415","created_at":"2026-07-05T05:27:38Z"},{"alias_kind":"arxiv_version","alias_value":"2212.11415v1","created_at":"2026-07-05T05:27:38Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2212.11415","created_at":"2026-07-05T05:27:38Z"},{"alias_kind":"pith_short_12","alias_value":"3RI5ZPJT67FO","created_at":"2026-07-05T05:27:38Z"},{"alias_kind":"pith_short_16","alias_value":"3RI5ZPJT67FOOWE5","created_at":"2026-07-05T05:27:38Z"},{"alias_kind":"pith_short_8","alias_value":"3RI5ZPJT","created_at":"2026-07-05T05:27:38Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2022:3RI5ZPJT67FOOWE5DHSCSYQKSK","target":"record","payload":{"canonical_record":{"source":{"id":"2212.11415","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2022-12-21T23:52:42Z","cross_cats_sorted":[],"title_canon_sha256":"cb357e916b270b668e8118ba71ca10d6ecbe55c66d6f38341ce06eca487a0869","abstract_canon_sha256":"fcd1d203779ff3901bd8c7b74f2254203ed30a98b325c2c1f4e144f24a8ca2a1"},"schema_version":"1.0"},"canonical_sha256":"dc51dcbd33f7cae7589d19e429620a929a0b5a07cd70296a3cb38887c361c050","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:27:38.823713Z","signature_b64":"fQyF2W13Her+ZeQVVP5P4zyyCqLEDfmfQ8KtotH7nCBihJFq/pXRiFLPPPtXj+yG4VB6fYETHPqcOisMzJAEDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"dc51dcbd33f7cae7589d19e429620a929a0b5a07cd70296a3cb38887c361c050","last_reissued_at":"2026-07-05T05:27:38.823297Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:27:38.823297Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2212.11415","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T05:27:38Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"RwNJuyVyttZBdmlIM3U0ZAGHM/kM5f6cagqThvj2Gl2h0k093y4T/NUK2on6KDNTHAK9k8JR49PFI1S5KUk/DQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-12T10:50:36.526298Z"},"content_sha256":"184e00ca7fd42539e39d2dcba3d2ea06f92210cc789ffc2d6a462c101e936470","schema_version":"1.0","event_id":"sha256:184e00ca7fd42539e39d2dcba3d2ea06f92210cc789ffc2d6a462c101e936470"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2022:3RI5ZPJT67FOOWE5DHSCSYQKSK","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Circumventing interpretability: How to defeat mind-readers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Lee Sharkey","submitted_at":"2022-12-21T23:52:42Z","abstract_excerpt":"The increasing capabilities of artificial intelligence (AI) systems make it ever more important that we interpret their internals to ensure that their intentions are aligned with human values. Yet there is reason to believe that misaligned artificial intelligence will have a convergent instrumental incentive to make its thoughts difficult for us to interpret. In this article, I discuss many ways that a capable AI might circumvent scalable interpretability methods and suggest a framework for thinking about these potential future risks."},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2212.11415","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/2212.11415/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T05:27:38Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"BAS3CXJPqvcwGWmbDzUeQ63EHHpGX/yJqtGj6u3V23Fjy8BzcNv9kBQ/w/dKNVQ6Ry5y8gUU7zKVHSJPbJfCDA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-12T10:50:36.527544Z"},"content_sha256":"1cf1f72d6a33c094751c10bbc404651c6847f313ab497b4c1342d935c3ccefeb","schema_version":"1.0","event_id":"sha256:1cf1f72d6a33c094751c10bbc404651c6847f313ab497b4c1342d935c3ccefeb"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/3RI5ZPJT67FOOWE5DHSCSYQKSK/bundle.json","state_url":"https://pith.science/pith/3RI5ZPJT67FOOWE5DHSCSYQKSK/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/3RI5ZPJT67FOOWE5DHSCSYQKSK/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-12T10:50:36Z","links":{"resolver":"https://pith.science/pith/3RI5ZPJT67FOOWE5DHSCSYQKSK","bundle":"https://pith.science/pith/3RI5ZPJT67FOOWE5DHSCSYQKSK/bundle.json","state":"https://pith.science/pith/3RI5ZPJT67FOOWE5DHSCSYQKSK/state.json","well_known_bundle":"https://pith.science/.well-known/pith/3RI5ZPJT67FOOWE5DHSCSYQKSK/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2022:3RI5ZPJT67FOOWE5DHSCSYQKSK","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"fcd1d203779ff3901bd8c7b74f2254203ed30a98b325c2c1f4e144f24a8ca2a1","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2022-12-21T23:52:42Z","title_canon_sha256":"cb357e916b270b668e8118ba71ca10d6ecbe55c66d6f38341ce06eca487a0869"},"schema_version":"1.0","source":{"id":"2212.11415","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2212.11415","created_at":"2026-07-05T05:27:38Z"},{"alias_kind":"arxiv_version","alias_value":"2212.11415v1","created_at":"2026-07-05T05:27:38Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2212.11415","created_at":"2026-07-05T05:27:38Z"},{"alias_kind":"pith_short_12","alias_value":"3RI5ZPJT67FO","created_at":"2026-07-05T05:27:38Z"},{"alias_kind":"pith_short_16","alias_value":"3RI5ZPJT67FOOWE5","created_at":"2026-07-05T05:27:38Z"},{"alias_kind":"pith_short_8","alias_value":"3RI5ZPJT","created_at":"2026-07-05T05:27:38Z"}],"graph_snapshots":[{"event_id":"sha256:1cf1f72d6a33c094751c10bbc404651c6847f313ab497b4c1342d935c3ccefeb","target":"graph","created_at":"2026-07-05T05:27:38Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2212.11415/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"The increasing capabilities of artificial intelligence (AI) systems make it ever more important that we interpret their internals to ensure that their intentions are aligned with human values. Yet there is reason to believe that misaligned artificial intelligence will have a convergent instrumental incentive to make its thoughts difficult for us to interpret. In this article, I discuss many ways that a capable AI might circumvent scalable interpretability methods and suggest a framework for thinking about these potential future risks.","authors_text":"Lee Sharkey","cross_cats":[],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2022-12-21T23:52:42Z","title":"Circumventing interpretability: How to defeat mind-readers"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2212.11415","kind":"arxiv","version":1},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:184e00ca7fd42539e39d2dcba3d2ea06f92210cc789ffc2d6a462c101e936470","target":"record","created_at":"2026-07-05T05:27:38Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"fcd1d203779ff3901bd8c7b74f2254203ed30a98b325c2c1f4e144f24a8ca2a1","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2022-12-21T23:52:42Z","title_canon_sha256":"cb357e916b270b668e8118ba71ca10d6ecbe55c66d6f38341ce06eca487a0869"},"schema_version":"1.0","source":{"id":"2212.11415","kind":"arxiv","version":1}},"canonical_sha256":"dc51dcbd33f7cae7589d19e429620a929a0b5a07cd70296a3cb38887c361c050","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"dc51dcbd33f7cae7589d19e429620a929a0b5a07cd70296a3cb38887c361c050","first_computed_at":"2026-07-05T05:27:38.823297Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T05:27:38.823297Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"fQyF2W13Her+ZeQVVP5P4zyyCqLEDfmfQ8KtotH7nCBihJFq/pXRiFLPPPtXj+yG4VB6fYETHPqcOisMzJAEDQ==","signature_status":"signed_v1","signed_at":"2026-07-05T05:27:38.823713Z","signed_message":"canonical_sha256_bytes"},"source_id":"2212.11415","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:184e00ca7fd42539e39d2dcba3d2ea06f92210cc789ffc2d6a462c101e936470","sha256:1cf1f72d6a33c094751c10bbc404651c6847f313ab497b4c1342d935c3ccefeb"],"state_sha256":"6ea6630a7bd33820e6b68f38c51f1943ad2db5aed6dcf83c13aef039f2d5d547"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"2HMtyzXdHeYxayXRuztBWo1XFfW9sx5I2SmlJD0dlsZ/WJtsfJmG1aDhdoxZ5JEv7flw1DsuMy58vm3LFZUiBg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-12T10:50:36.537323Z","bundle_sha256":"2c4f984394ae2ee9f2641661de10cba75c6980e4c2941729c3e3e188a4c0fe18"}}