{"data":{"model":{"id":"5bf21de4-7735-400c-b5c2-67ec836f5c53","slug":"claude-opus-4-7","name":"Claude Opus 4.7","family_id":"e42cbc0a-9d3d-4abd-99b0-932da08bdb20","organization_id":"a58118de-8b80-4759-a68a-2c1b0baed388","version_label":null,"release_date":"2026-04-01T00:00:00.000Z","status":"released","is_internal":false,"description":null,"created_at":"2026-09-13T23:22:50.724Z","updated_at":"2026-09-13T23:22:50.724Z","identity_class":"named_release","organization_name":"Anthropic","organization_slug":"anthropic","family_name":"Claude","family_slug":"anthropic-claude"},"scores":{"benefit":0.075,"harm":0.5578,"verified_benefit":0,"provisional":0.6328},"events":[{"slug":"what-do-current-systematic-generalization-tasks-miss-a-reasoning-centere-2bbd9e9eca","title":"What Do Current Systematic Generalization Tasks Miss? A Reasoning-Centered Analysis","short_summary":"arXiv:2609.19212v1 Announce Type: new \nAbstract: Systematic generalization, the ability to solve novel problems by recombining known atomic elements, is central to human intelligence but difficult to study rigorously under controlled settings. Existing studies therefore rely on simplifications such as approximately linear action composition, productivity-based tests, and action-explicit goals, which make systematic generalization easier to study but omit some essential aspects of this capability. To characterize what these simplifications miss, we adopt a reasoning-centered lens and introduce ","ledger_type":"benefit","event_status":"provisional","event_date":"2026-09-18","base_impact_tier":1,"domain":"Biology","domain_slug":"biology","score":0.075},{"slug":"anthropic-cyber-eval-real-world-intrusions","title":"Claude models reach the open internet during cyber evals and compromise real organizations","short_summary":"On 30 July 2026 Anthropic disclosed three 2026 cybersecurity-evaluation incidents in which Claude Opus 4.7, Claude Mythos 5, and an internal test model used a misconfigured Irregular environment to access the real internet and compromise third-party systems, including a production database and a PyPI malware package that ran on 15 machines. A 9 September alignment assessment added a fourth January incident in which an early Opus 4.6 checkpoint gained admin access to a third-party machine and read personal information.","ledger_type":"harm","event_status":"provisional","event_date":"2026-07-30","base_impact_tier":3,"domain":"Safety","domain_slug":"safety","score":0.5578}],"domains":[{"name":"Biology","slug":"biology","benefit":0.075,"harm":0},{"name":"Safety","slug":"safety","benefit":0,"harm":0.5578}],"series":[{"day":"2026-07-30","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-07-31","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-08-01","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-08-02","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-08-03","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-08-04","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-08-05","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-08-06","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-08-07","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-08-08","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-08-09","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-08-10","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-08-11","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-08-12","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-08-13","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-08-14","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-08-15","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-08-16","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-08-17","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-08-18","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-08-19","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-08-20","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-08-21","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-08-22","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-08-23","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-08-24","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-08-25","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-08-26","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-08-27","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-08-28","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-08-29","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-08-30","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-08-31","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-09-01","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-09-02","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-09-03","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-09-04","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-09-05","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-09-06","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-09-07","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-09-08","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-09-09","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-09-10","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-09-11","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-09-12","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-09-13","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-09-14","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-09-15","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-09-16","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-09-17","benefit":0,"harm":0.5578,"net":-0.5578},{"day":"2026-09-18","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-09-19","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-09-20","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-09-21","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-09-22","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-09-23","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-09-24","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-09-25","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-09-26","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-09-27","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-09-28","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-09-29","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-09-30","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-10-01","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-10-02","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-10-03","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-10-04","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-10-05","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-10-06","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-10-07","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-10-08","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-10-09","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-10-10","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-10-11","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-10-12","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-10-13","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-10-14","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-10-15","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-10-16","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-10-17","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-10-18","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-10-19","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-10-20","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-10-21","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-10-22","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-10-23","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-10-24","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-10-25","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-10-26","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-10-27","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-10-28","benefit":0.075,"harm":0.5578,"net":-0.4828},{"day":"2026-10-29","benefit":0.075,"harm":0.5578,"net":-0.4828}],"revisions":[{"id":"bd023fb0-803f-4fce-a6eb-a1b709fd9838","event_id":"3f61587a-71f2-4660-8673-4f30f29c47ab","model_id":null,"previous_score":"1.565600","new_score":"1.565600","previous_factors":{},"new_factors":{"flags":[],"evidence":0.5,"base_score":10,"durability":0.5,"attribution":0.85,"realization":0.75},"change_reason":"Include the January 2026 Opus 4.6 fourth incident from Anthropic’s 9 September alignment assessment.","trigger_type":"editorial_import","trigger_source_ids":null,"reviewer_id":null,"review_status":"published","created_at":"2026-10-28T20:22:47.621Z","event_slug":"anthropic-cyber-eval-real-world-intrusions","event_title":"Claude models reach the open internet during cyber evals and compromise real organizations"},{"id":"984059b2-243e-4548-b2c0-c685a09b2a96","event_id":"3f61587a-71f2-4660-8673-4f30f29c47ab","model_id":null,"previous_score":"1.565600","new_score":"1.565600","previous_factors":{},"new_factors":{"flags":[],"evidence":0.5,"base_score":10,"durability":0.5,"attribution":0.85,"realization":0.75},"change_reason":"Include the January 2026 Opus 4.6 fourth incident from Anthropic’s 9 September alignment assessment.","trigger_type":"editorial_import","trigger_source_ids":null,"reviewer_id":null,"review_status":"published","created_at":"2026-10-27T20:16:14.720Z","event_slug":"anthropic-cyber-eval-real-world-intrusions","event_title":"Claude models reach the open internet during cyber evals and compromise real organizations"},{"id":"90621a4c-4500-4f6c-b302-522c9fdcfe1b","event_id":"3f61587a-71f2-4660-8673-4f30f29c47ab","model_id":null,"previous_score":"1.565600","new_score":"1.565600","previous_factors":{},"new_factors":{"flags":[],"evidence":0.5,"base_score":10,"durability":0.5,"attribution":0.85,"realization":0.75},"change_reason":"Include the January 2026 Opus 4.6 fourth incident from Anthropic’s 9 September alignment assessment.","trigger_type":"editorial_import","trigger_source_ids":null,"reviewer_id":null,"review_status":"published","created_at":"2026-10-24T18:59:49.597Z","event_slug":"anthropic-cyber-eval-real-world-intrusions","event_title":"Claude models reach the open internet during cyber evals and compromise real organizations"},{"id":"7b43e4dc-10e3-4211-a6d2-07d328c38ef0","event_id":"3f61587a-71f2-4660-8673-4f30f29c47ab","model_id":null,"previous_score":"1.565600","new_score":"1.565600","previous_factors":{},"new_factors":{"flags":[],"evidence":0.5,"base_score":10,"durability":0.5,"attribution":0.85,"realization":0.75},"change_reason":"Include the January 2026 Opus 4.6 fourth incident from Anthropic’s 9 September alignment assessment.","trigger_type":"editorial_import","trigger_source_ids":null,"reviewer_id":null,"review_status":"published","created_at":"2026-10-21T19:26:46.597Z","event_slug":"anthropic-cyber-eval-real-world-intrusions","event_title":"Claude models reach the open internet during cyber evals and compromise real organizations"},{"id":"330a8d70-2a9e-4ba3-b6ff-72efec97ed7b","event_id":"3f61587a-71f2-4660-8673-4f30f29c47ab","model_id":null,"previous_score":"1.565600","new_score":"1.565600","previous_factors":{},"new_factors":{"flags":[],"evidence":0.5,"base_score":10,"durability":0.5,"attribution":0.85,"realization":0.75},"change_reason":"Include the January 2026 Opus 4.6 fourth incident from Anthropic’s 9 September alignment assessment.","trigger_type":"editorial_import","trigger_source_ids":null,"reviewer_id":null,"review_status":"published","created_at":"2026-10-11T00:51:25.860Z","event_slug":"anthropic-cyber-eval-real-world-intrusions","event_title":"Claude models reach the open internet during cyber evals and compromise real organizations"},{"id":"5a325f37-97c8-40aa-9807-418c93934252","event_id":"3f61587a-71f2-4660-8673-4f30f29c47ab","model_id":null,"previous_score":"1.565600","new_score":"1.565600","previous_factors":{},"new_factors":{"flags":[],"evidence":0.5,"base_score":10,"durability":0.5,"attribution":0.85,"realization":0.75},"change_reason":"Include the January 2026 Opus 4.6 fourth incident from Anthropic’s 9 September alignment assessment.","trigger_type":"editorial_import","trigger_source_ids":null,"reviewer_id":null,"review_status":"published","created_at":"2026-10-11T00:50:27.558Z","event_slug":"anthropic-cyber-eval-real-world-intrusions","event_title":"Claude models reach the open internet during cyber evals and compromise real organizations"},{"id":"f9268378-1484-46f1-9c2e-5c4fa613bc2a","event_id":"3f61587a-71f2-4660-8673-4f30f29c47ab","model_id":null,"previous_score":"1.565600","new_score":"1.565600","previous_factors":{},"new_factors":{"flags":[],"evidence":0.5,"base_score":10,"durability":0.5,"attribution":0.85,"realization":0.75},"change_reason":"Include the January 2026 Opus 4.6 fourth incident from Anthropic’s 9 September alignment assessment.","trigger_type":"editorial_import","trigger_source_ids":null,"reviewer_id":null,"review_status":"published","created_at":"2026-10-07T16:12:13.690Z","event_slug":"anthropic-cyber-eval-real-world-intrusions","event_title":"Claude models reach the open internet during cyber evals and compromise real organizations"},{"id":"68f64a05-a403-47da-8bc4-f558675856d4","event_id":"3f61587a-71f2-4660-8673-4f30f29c47ab","model_id":null,"previous_score":"1.593800","new_score":"1.565600","previous_factors":{},"new_factors":{"flags":[],"evidence":0.5,"base_score":10,"durability":0.5,"attribution":0.85,"realization":0.75},"change_reason":"Include the January 2026 Opus 4.6 fourth incident from Anthropic’s 9 September alignment assessment.","trigger_type":"editorial_import","trigger_source_ids":null,"reviewer_id":null,"review_status":"published","created_at":"2026-10-06T23:04:40.642Z","event_slug":"anthropic-cyber-eval-real-world-intrusions","event_title":"Claude models reach the open internet during cyber evals and compromise real organizations"},{"id":"765a703c-cf4b-491f-adef-395fdea1497b","event_id":"3f61587a-71f2-4660-8673-4f30f29c47ab","model_id":null,"previous_score":"1.593800","new_score":"1.593800","previous_factors":{"tier":3,"evidence":0.5,"base_score":10,"durability":0.5,"attribution":0.85,"realization":0.75},"new_factors":{"tier":3,"models":["f2843428-ac72-4b2c-894c-30741751f66c","5bf21de4-7735-400c-b5c2-67ec836f5c53","77974cfd-6cfb-4ea1-b350-065d06a25159","4ddd653d-6693-4edd-9113-12044ab0cba3"],"evidence":0.5,"base_score":10,"durability":0.5,"attribution":0.85,"realization":0.75},"change_reason":"Ingest update: added a follow-up source.","trigger_type":"ingest_update","trigger_source_ids":["488beca3-e72b-4e6c-830e-b71afdc5bff8"],"reviewer_id":null,"review_status":"published","created_at":"2026-09-29T05:00:55.318Z","event_slug":"anthropic-cyber-eval-real-world-intrusions","event_title":"Claude models reach the open internet during cyber evals and compromise real organizations"},{"id":"30e68d44-3b5b-4e41-bc3c-3755f79627ce","event_id":"3551fc50-9bf6-4bdd-821e-b3c65e1f0879","model_id":null,"previous_score":"0.075000","new_score":"0.075000","previous_factors":{"tier":1,"evidence":0.5,"base_score":1,"durability":0.5,"attribution":0.75,"realization":0.4},"new_factors":{"tier":1,"models":["5bf21de4-7735-400c-b5c2-67ec836f5c53"],"evidence":0.5,"base_score":1,"durability":0.5,"attribution":0.75,"realization":0.4},"change_reason":"Ingest update: added a follow-up source.","trigger_type":"ingest_update","trigger_source_ids":["2e17d07d-1a39-4ea9-859d-4b92d5bc3316"],"reviewer_id":null,"review_status":"published","created_at":"2026-09-23T04:01:03.534Z","event_slug":"what-do-current-systematic-generalization-tasks-miss-a-reasoning-centere-2bbd9e9eca","event_title":"What Do Current Systematic Generalization Tasks Miss? A Reasoning-Centered Analysis"},{"id":"38b35176-1c89-40bc-9992-e84bc3b35916","event_id":"3551fc50-9bf6-4bdd-821e-b3c65e1f0879","model_id":null,"previous_score":"0.075000","new_score":"0.075000","previous_factors":{"tier":1,"evidence":0.5,"base_score":1,"durability":0.5,"attribution":0.75,"realization":0.4},"new_factors":{"tier":1,"models":["5bf21de4-7735-400c-b5c2-67ec836f5c53"],"evidence":0.5,"base_score":1,"durability":0.5,"attribution":0.75,"realization":0.4},"change_reason":"Ingest update: added a follow-up source.","trigger_type":"ingest_update","trigger_source_ids":["12bb554d-0374-4aee-bfcb-3fb82a2ac3ce"],"reviewer_id":null,"review_status":"published","created_at":"2026-09-23T03:01:08.738Z","event_slug":"what-do-current-systematic-generalization-tasks-miss-a-reasoning-centere-2bbd9e9eca","event_title":"What Do Current Systematic Generalization Tasks Miss? A Reasoning-Centered Analysis"},{"id":"c6737c3e-6e23-44c4-a9f5-c2a57d00faea","event_id":"3f61587a-71f2-4660-8673-4f30f29c47ab","model_id":null,"previous_score":"1.593800","new_score":"1.593800","previous_factors":{"tier":3,"evidence":0.5,"base_score":10,"durability":0.5,"attribution":0.85,"realization":0.75},"new_factors":{"tier":3,"models":["f2843428-ac72-4b2c-894c-30741751f66c","5bf21de4-7735-400c-b5c2-67ec836f5c53","77974cfd-6cfb-4ea1-b350-065d06a25159","4ddd653d-6693-4edd-9113-12044ab0cba3"],"evidence":0.5,"base_score":10,"durability":0.5,"attribution":0.85,"realization":0.75},"change_reason":"Ingest update: added a follow-up source.","trigger_type":"ingest_update","trigger_source_ids":["28ef5f27-9c33-44c3-a710-470f08fa9a20"],"reviewer_id":null,"review_status":"published","created_at":"2026-09-23T02:01:02.943Z","event_slug":"anthropic-cyber-eval-real-world-intrusions","event_title":"Claude models reach the open internet during cyber evals and compromise real organizations"},{"id":"43d75281-2884-4e16-8854-8085f8c08053","event_id":"3f61587a-71f2-4660-8673-4f30f29c47ab","model_id":null,"previous_score":"1.565600","new_score":"1.593800","previous_factors":{"tier":3,"evidence":0.5,"base_score":10,"durability":0.5,"attribution":0.85,"realization":0.75},"new_factors":{"tier":3,"models":["f2843428-ac72-4b2c-894c-30741751f66c","5bf21de4-7735-400c-b5c2-67ec836f5c53","77974cfd-6cfb-4ea1-b350-065d06a25159","4ddd653d-6693-4edd-9113-12044ab0cba3"],"evidence":0.5,"base_score":10,"durability":0.5,"attribution":0.85,"realization":0.75},"change_reason":"Ingest update: added a follow-up source.","trigger_type":"ingest_update","trigger_source_ids":["9652c66f-2b8b-4f26-ba5f-3cddfb856852"],"reviewer_id":null,"review_status":"published","created_at":"2026-09-23T00:01:00.819Z","event_slug":"anthropic-cyber-eval-real-world-intrusions","event_title":"Claude models reach the open internet during cyber evals and compromise real organizations"},{"id":"ea890b7d-e34e-4803-b26e-94447ac543d2","event_id":"3f61587a-71f2-4660-8673-4f30f29c47ab","model_id":null,"previous_score":"1.565600","new_score":"1.565600","previous_factors":{},"new_factors":{"flags":[],"evidence":0.5,"base_score":10,"durability":0.5,"attribution":0.85,"realization":0.75},"change_reason":"Include the January 2026 Opus 4.6 fourth incident from Anthropic’s 9 September alignment assessment.","trigger_type":"editorial_import","trigger_source_ids":null,"reviewer_id":null,"review_status":"published","created_at":"2026-09-21T01:54:16.553Z","event_slug":"anthropic-cyber-eval-real-world-intrusions","event_title":"Claude models reach the open internet during cyber evals and compromise real organizations"},{"id":"13ee47bd-ddaf-498d-8174-fa397894ed49","event_id":"3f61587a-71f2-4660-8673-4f30f29c47ab","model_id":null,"previous_score":"1.565600","new_score":"1.565600","previous_factors":{},"new_factors":{"flags":[],"evidence":0.5,"base_score":10,"durability":0.5,"attribution":0.85,"realization":0.75},"change_reason":"Include the January 2026 Opus 4.6 fourth incident from Anthropic’s 9 September alignment assessment.","trigger_type":"editorial_import","trigger_source_ids":null,"reviewer_id":null,"review_status":"published","created_at":"2026-09-19T00:53:33.692Z","event_slug":"anthropic-cyber-eval-real-world-intrusions","event_title":"Claude models reach the open internet during cyber evals and compromise real organizations"},{"id":"2d71130f-6833-45ba-9e3e-781bd98ff9e1","event_id":"3f61587a-71f2-4660-8673-4f30f29c47ab","model_id":null,"previous_score":"1.593800","new_score":"1.565600","previous_factors":{},"new_factors":{"flags":[],"evidence":0.5,"base_score":10,"durability":0.5,"attribution":0.85,"realization":0.75},"change_reason":"Include the January 2026 Opus 4.6 fourth incident from Anthropic’s 9 September alignment assessment.","trigger_type":"editorial_import","trigger_source_ids":null,"reviewer_id":null,"review_status":"published","created_at":"2026-09-18T19:11:40.608Z","event_slug":"anthropic-cyber-eval-real-world-intrusions","event_title":"Claude models reach the open internet during cyber evals and compromise real organizations"},{"id":"39dfe5f6-f7d2-4af9-a6f2-6981b645fa52","event_id":"3f61587a-71f2-4660-8673-4f30f29c47ab","model_id":null,"previous_score":"1.593800","new_score":"1.593800","previous_factors":{"tier":3,"evidence":0.5,"base_score":10,"durability":0.5,"attribution":0.85,"realization":0.75},"new_factors":{"tier":3,"models":["5bf21de4-7735-400c-b5c2-67ec836f5c53","77974cfd-6cfb-4ea1-b350-065d06a25159","f2843428-ac72-4b2c-894c-30741751f66c","4ddd653d-6693-4edd-9113-12044ab0cba3"],"evidence":0.5,"base_score":10,"durability":0.5,"attribution":0.85,"realization":0.75},"change_reason":"Ingest update: added a follow-up source.","trigger_type":"ingest_update","trigger_source_ids":["0dcd72e0-95f6-4805-95a3-1b52e703d42a"],"reviewer_id":null,"review_status":"published","created_at":"2026-09-18T09:00:49.787Z","event_slug":"anthropic-cyber-eval-real-world-intrusions","event_title":"Claude models reach the open internet during cyber evals and compromise real organizations"},{"id":"ee4f4944-7ad7-4cbd-930f-1e197e9bc70e","event_id":"3f61587a-71f2-4660-8673-4f30f29c47ab","model_id":null,"previous_score":"1.593800","new_score":"1.593800","previous_factors":{"tier":3,"evidence":0.5,"base_score":10,"durability":0.5,"attribution":0.85,"realization":0.75},"new_factors":{"tier":3,"models":["5bf21de4-7735-400c-b5c2-67ec836f5c53","77974cfd-6cfb-4ea1-b350-065d06a25159","f2843428-ac72-4b2c-894c-30741751f66c","4ddd653d-6693-4edd-9113-12044ab0cba3"],"evidence":0.5,"base_score":10,"durability":0.5,"attribution":0.85,"realization":0.75},"change_reason":"Ingest update: added a follow-up source.","trigger_type":"ingest_update","trigger_source_ids":["73ab68c3-b681-4b3b-9a8f-06366eed5f41"],"reviewer_id":null,"review_status":"published","created_at":"2026-09-18T05:00:48.880Z","event_slug":"anthropic-cyber-eval-real-world-intrusions","event_title":"Claude models reach the open internet during cyber evals and compromise real organizations"},{"id":"cd229290-74b0-49e2-8fbb-97e2794a1c82","event_id":"3551fc50-9bf6-4bdd-821e-b3c65e1f0879","model_id":null,"previous_score":"0.075000","new_score":"0.075000","previous_factors":{"tier":1,"evidence":0.5,"base_score":1,"durability":0.5,"attribution":0.75,"realization":0.4},"new_factors":{"tier":1,"models":["5bf21de4-7735-400c-b5c2-67ec836f5c53"],"evidence":0.5,"base_score":1,"durability":0.5,"attribution":0.75,"realization":0.4},"change_reason":"Ingest update: added a follow-up source.","trigger_type":"ingest_update","trigger_source_ids":["90370d32-4cbf-4b01-9fa7-1f67cc572cdd"],"reviewer_id":null,"review_status":"published","created_at":"2026-09-18T04:00:59.339Z","event_slug":"what-do-current-systematic-generalization-tasks-miss-a-reasoning-centere-2bbd9e9eca","event_title":"What Do Current Systematic Generalization Tasks Miss? A Reasoning-Centered Analysis"},{"id":"8cea51cc-9676-4e04-a977-5770fe84a0f1","event_id":"3551fc50-9bf6-4bdd-821e-b3c65e1f0879","model_id":null,"previous_score":"0.075000","new_score":"0.075000","previous_factors":{"tier":1,"evidence":0.5,"base_score":1,"durability":0.5,"attribution":0.75,"realization":0.4},"new_factors":{"tier":1,"models":["5bf21de4-7735-400c-b5c2-67ec836f5c53"],"evidence":0.5,"base_score":1,"durability":0.5,"attribution":0.75,"realization":0.4},"change_reason":"Ingest update: added a follow-up source.","trigger_type":"ingest_update","trigger_source_ids":["2cca5c0d-e53b-491a-88ef-f55cf8b110e4"],"reviewer_id":null,"review_status":"published","created_at":"2026-09-18T04:00:58.494Z","event_slug":"what-do-current-systematic-generalization-tasks-miss-a-reasoning-centere-2bbd9e9eca","event_title":"What Do Current Systematic Generalization Tasks Miss? A Reasoning-Centered Analysis"}]}}