{"schema_version":"0.10.0","id":"LR-2026-0528","record_version":1,"title":"Claude Mythos accessed by unauthorized group via leaked naming and contractor login","verification":{"human_verified":false,"method":"automated_gate","auto_verified":true,"score":59.9,"score_breakdown":{"source_authority":12.1,"independent_corroboration":15.0,"implicated_party_confirmed":0.0,"citation_integrity":14.0,"specificity":6.0,"completeness":4.8,"unhedged_language":5.0},"checks":{"schema_valid":true,"every_field_cited":true,"quotes_verbatim":true,"source_fetched":true,"distinct_incident":true,"dates_plausible":true,"sources_agree":true,"source_describes_incident":true,"severe_backed_by_source":true,"record_integrity":true,"eligible":true,"about_this_case":true},"gate_version":"gate-2026-10-18.7","cross_check_model":null,"cross_check_agreement":null,"duplicate_of":null,"verified_by":null,"verified_at":"2026-11-01","extraction_run_id":null,"evidence_level":{"level":"preliminary","reason":"Publisher not yet named · subject to change","original_document":null,"reports_total":2,"reports_independent":0,"reports_derivative":0,"developing":false,"disputed":false,"rules_version":"evidence-2026-10-29.1","computed_at":"2026-11-01","history":[{"date":"2026-11-01","from_level":null,"to_level":"preliminary","reason":"Publisher not yet named · subject to change"}]},"sources_hash":"7351121f71fb3f27137ee6acf5577a80d0db6a90fafa37ecef8e3808a5defc7e"},"system":{"model_name":"Claude Mythos","model_version":null,"developer":"Anthropic","harness":null,"tools_granted":[],"permissions_scope":null,"autonomy_level":"unknown","guardrails_present":[],"deployment_context":"unknown","tier":null,"lifecycle_stage":null,"integration_type":null,"intended_use":null,"safeguard_state":null,"production_intended":null},"event":{"date_occurred":null,"date_disclosed":"2026-04-23","incident_class":"security_compromise","failure_class":null,"agent_involved":null,"secondary_failure_classes":[],"trigger":"credential_exposure","duration":null,"detection_method":"third_party","summary":"Anthropic’s Claude Mythos, a model offered to select partners for testing, was reportedly accessed without authorization. Bloomberg reported that a small group of unauthorized users gained access by guessing the model’s online location using information exposed in the Mercor breach, along with a contractor’s access. Bloomberg reported that the group has been using Mythos since release day. The group told Bloomberg it did not use Mythos for cyberattacks. A reporter uncovered the breach rather than Anthropic. Anthropic says it is investigating. Security commentators said the access route was predictable and that monitoring should have spotted it.","tags":[],"date_detected":null,"near_miss":null,"root_cause":null,"root_cause_detail":null,"failure_class_detail":null,"responsible_party":null,"adversarial_actor":"external_attacker","input_provenance":null,"time_to_contain":null,"operations_paused":null,"remediation":null,"precursor_events":null,"primary_risk":"RK03","secondary_risks":["RK08"]},"impact":{"direct_loss_usd":null,"remediation_hours":null,"records_exposed":null,"third_parties_affected_count":null,"severity":"moderate","severity_because":"Moderate test: systems of others touched with a bounded scope stated, since the access was by a small group of unauthorized users.","harm_class":[],"affected_description":"Access to the Mythos model by a small group of unauthorized users","protected_class_involved":null,"data_categories_exposed":[],"loss_range_usd":null,"reversibility":null,"downstream_exposure_count":null},"evidence":{"sources":[{"url":"https://www.theverge.com/ai-artificial-intelligence/917644/anthropic-claude-mythos-breach-humiliation","publisher":"The Verge","date":"2026-04-23","type":"press","title":"Anthropic’s Mythos breach was humiliating - The Verge","archived_url":null,"attachment":"first_report"},{"url":"https://www.therundown.ai/articles/anthropic-locked-down-mythos-leaks","publisher":"The Rundown AI","date":"2026-04-23","type":"press","title":"Anthropic's locked-down Mythos leaks - The Rundown AI","archived_url":null}],"per_field_confidence":{"system.model_name":"C","system.developer":"C","event.incident_class":"C","event.trigger":"C","event.detection_method":"C","event.adversarial_actor":"C","impact.affected_description":"C","context.industry":"C","impact.severity":"D","impact.severity_because":"D","event.primary_risk":"D","event.secondary_risks":"D","event.summary":"C","system.autonomy_level":"D","system.deployment_context":"D","context.country":"D","event.date_disclosed":"D","context.denominator_available":"D"},"field_citations":{"system.model_name":[{"source_index":0,"quote":"Anthropic’s tightly controlled rollout of Claude Mythos has taken an awkward turn."}],"system.developer":[{"source_index":0,"quote":"Anthropic’s tightly controlled rollout of Claude Mythos has taken an awkward turn."}],"event.incident_class":[{"source_index":0,"quote":"According to Bloomberg, a “small group of unauthorized users” has had access to Mythos — whose existence was first revealed in a leak — since the day Anthropic announced plans to offer it to a select group of companies for testing."}],"event.trigger":[{"source_index":1,"quote":"One member had vendor credentials through contract work, with leaked Mercor details helping the group locate and access Mythos online."}],"event.detection_method":[{"source_index":0,"quote":"The fact that the breach was uncovered by a reporter rather than Anthropic also raises the obvious question of whether it is an isolated incident."}],"event.adversarial_actor":[{"source_index":0,"quote":"According to Bloomberg, a “small group of unauthorized users” has had access to Mythos — whose existence was first revealed in a leak — since the day Anthropic announced plans to offer it to a select group of companies for testing."}],"impact.affected_description":[{"source_index":0,"quote":"According to Bloomberg, a “small group of unauthorized users” has had access to Mythos — whose existence was first revealed in a leak — since the day Anthropic announced plans to offer it to a select group of companies for testing."}],"context.industry":[{"source_index":0,"quote":"Anthropic’s tightly controlled rollout of Claude Mythos has taken an awkward turn."}],"impact.severity":[{"source_index":0,"quote":"According to Bloomberg, a “small group of unauthorized users” has had access to Mythos — whose existence was first revealed in a leak — since the day Anthropic announced plans to offer it to a select group of companies for testing."}],"impact.severity_because":[{"source_index":0,"quote":"According to Bloomberg, a “small group of unauthorized users” has had access to Mythos — whose existence was first revealed in a leak — since the day Anthropic announced plans to offer it to a select group of companies for testing."}],"event.summary":[{"source_index":0,"quote":"According to Bloomberg, a “small group of unauthorized users” has had access to Mythos — whose existence was first revealed in a leak — since the day Anthropic announced plans to offer it to a select group of companies for testing."},{"source_index":0,"quote":"Bloomberg reports the group accessed Mythos by making “an educated guess about the model’s online location,” using information about Anthropic’s other models exposed in the breach of Mercor — a company that makes AI training data — along with access one member had through contract work evaluating Anthropic models."},{"source_index":1,"quote":"Bloomberg reported that a private Discord group tracking unreleased models accessed Mythos on release day and has been using it regularly."},{"source_index":1,"quote":"The group told Bloomberg that they do not use Mythos for cyberattacks or other malicious activities, and also claimed access to other unreleased models."},{"source_index":0,"quote":"The fact that the breach was uncovered by a reporter rather than Anthropic also raises the obvious question of whether it is an isolated incident."},{"source_index":0,"quote":"Anthropic says it is investigating."},{"source_index":0,"quote":"Evidently, Anthropic wasn’t monitoring closely enough — and given how dangerous it says the model is, it’s reasonable to ask why."}]},"notes_on_gaps":"Sources 0 and 1 are press reports (The Verge citing Bloomberg; The Rundown citing Bloomberg). No statement from Anthropic beyond it is investigating, and no model version, harness, tools, permissions, autonomy level, lifecycle stage or deployment context is stated, so those fields are omitted. No containment time, duration, remediation, records or loss figures are stated. Year 2026 for the April 10 release is taken from the source date headers, not the quoted sentence. date_disclosed uses the article date of Source 0, a judgement. date_detected is omitted because the sources give no detection date. Trigger credential_exposure and root_cause integration are my classification of the contractor credential and leaked Mercor details. Adversarial_actor external_attacker is my judgement, since the group is described as unauthorized users, while one member had insider contract access. Responsible_party is omitted as no source states it. Severity moderate is a judgement that the affected systems are Anthropic's own and the scope is bounded by a small group. Control outcomes are my reading of the press descriptions of monitoring and patching. Source 0 contains journalistic opinion, which I did not use as a finding. [case created by the backfill: read by claude-haiku-5-5, backfill-2026-11-01.2, 2026-11-01; not yet reviewed] [2026-11-01: evidence level set: Preliminary (Publisher not yet named · subject to change)]"},"context":{"industry":"artificial intelligence","org_size_band":null,"country":"unknown","denominator_available":false,"denominator":null,"related_incident_ids":[],"systemic_exposure":null,"shared_model_id":null,"upstream_incident_id":null,"incident_evidence":null,"regulatory":null,"coverage":null},"narrative":{"steps":[{"step":"doing","text":"Mythos was being offered to a select group of companies for testing under a restricted rollout.","quote":"According to Bloomberg, a “small group of unauthorized users” has had access to Mythos — whose existence was first revealed in a leak — since the day Anthropic announced plans to offer it to a select group of companies for testing.","source_index":0,"grade":"C"},{"step":"went_wrong","text":"A small group guessed the model's location using information exposed in the Mercor breach and a contractor's access.","quote":"Bloomberg reports the group accessed Mythos by making “an educated guess about the model’s online location,” using information about Anthropic’s other models exposed in the breach of Mercor — a company that makes AI training data — along with access one member had through contract work evaluating Anthropic models.","source_index":0,"grade":"C"},{"step":"reached","text":"The group reportedly gained access to Mythos using leaked naming conventions and a borrowed contractor login.","quote":"Working off naming conventions leaked in a data breach and a borrowed contractor login, the group slipped into Anthropic's most restricted AI to date — a bad start for keeping increasingly powerful systems under wraps.","source_index":1,"grade":"C"},{"step":"caught","text":"A reporter uncovered the breach rather than Anthropic.","quote":"The fact that the breach was uncovered by a reporter rather than Anthropic also raises the obvious question of whether it is an isolated incident.","source_index":0,"grade":"C"}],"model":"claude-haiku-5-5","prompt_version":"backfill-2026-11-01.2","extracted_at":"2026-11-01","source_urls":["https://www.theverge.com/ai-artificial-intelligence/917644/anthropic-claude-mythos-breach-humiliation","https://www.therundown.ai/articles/anthropic-locked-down-mythos-leaks"]},"analysis":{"sections":{"timeline":[{"text":"Existence of the model accidentally revealed before release through an unsecured data trove.","quote":"The model’s existence was accidentally revealed before release through an “unsecured data trove” on a central system containing content for its website.","source_index":0,"when":null},{"text":"Mythos released to select partners under Project Glasswing.","quote":"The cybersecurity model was released on April 10 to select partners under ‘Project Glasswing’, with Anthropic deeming it too powerful for public release.","source_index":1,"when":"April 10"},{"text":"A private Discord group accessed Mythos on release day and kept using it regularly.","quote":"Bloomberg reported that a private Discord group tracking unreleased models accessed Mythos on release day and has been using it regularly.","source_index":1,"when":"release day"},{"text":"A reporter uncovered the breach rather than Anthropic.","quote":"The fact that the breach was uncovered by a reporter rather than Anthropic also raises the obvious question of whether it is an isolated incident.","source_index":0,"when":null},{"text":"Anthropic said it was investigating.","quote":"Anthropic says it is investigating.","source_index":0,"when":null}]},"model":"claude-haiku-5-5","prompt_version":"backfill-2026-11-01.2","extracted_at":"2026-11-01","source_urls":["https://www.theverge.com/ai-artificial-intelligence/917644/anthropic-claude-mythos-breach-humiliation","https://www.therundown.ai/articles/anthropic-locked-down-mythos-leaks"]}}