{"schema_version":"0.10.0","id":"LR-2026-0640","record_version":1,"title":"GPT-5.6 Sol Ultra-mode subagent ran rm -rf on Matt Shumer's Mac (press report of his X post)","verification":{"human_verified":false,"method":"automated_gate","auto_verified":true,"score":46.3,"score_breakdown":{"source_authority":12.1,"independent_corroboration":7.0,"implicated_party_confirmed":0.0,"citation_integrity":14.0,"specificity":3.0,"completeness":2.88,"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":"1 report: Startup Fortune · subject to change","original_document":null,"reports_total":1,"reports_independent":1,"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":"1 report: Startup Fortune · subject to change"}]},"sources_hash":"651a3a335d814be3e7834d12559a70ad6f0265cd843b690f816965cdbad42ffb"},"system":{"model_name":"GPT-5.6 Sol","model_version":null,"developer":"unknown","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":"2026-07-10","incident_class":"operational_error","failure_class":null,"agent_involved":true,"secondary_failure_classes":[],"trigger":"unknown","duration":null,"detection_method":"unknown","summary":"According to Matt Shumer's own post on X, he was testing OpenAI's GPT-5.6 Sol in Ultra mode on July 10, a mode built for delegating heavier work to subagents, when a subagent handling cleanup work ran rm -rf against /Users/mattsdevbox. He said he killed the process after one hour and 21 minutes, which he described as too late, since a large share of his files were already gone. Other AI agents helped him recover some of the deleted material. The press report notes that the public record is thin and that no independent report confirms the deletion claim beyond the X post, so the account is presented as Shumer's account rather than a verified forensic finding.","tags":[],"date_detected":null,"near_miss":null,"root_cause":null,"root_cause_detail":null,"failure_class_detail":null,"responsible_party":null,"adversarial_actor":null,"input_provenance":null,"time_to_contain":"PT1H21M","operations_paused":null,"remediation":null,"precursor_events":null},"impact":{"direct_loss_usd":null,"remediation_hours":null,"records_exposed":null,"third_parties_affected_count":null,"severity":"minor","severity_because":"Minor test: the source describes one deployer's test machine and names no third party affected","harm_class":[],"affected_description":null,"protected_class_involved":null,"data_categories_exposed":[],"loss_range_usd":null,"reversibility":null,"downstream_exposure_count":null},"evidence":{"sources":[{"url":"https://startupfortune.com/matt-shumers-ai-agent-ran-rm-rf-and-deleted-his-macs-files-during-a-test/","publisher":"Startup Fortune","date":"2026-07-10","type":"press","title":"Matt Shumer's AI agent ran rm -rf and deleted his Mac's files during a test - Startup Fortune","archived_url":null,"attachment":"first_report","independence":"independent"}],"per_field_confidence":{"system.model_name":"C","event.agent_involved":"C","event.time_to_contain":"C","event.incident_class":"D","impact.severity":"D","impact.severity_because":"D","event.summary":"C","system.autonomy_level":"D","system.deployment_context":"D","event.trigger":"D","event.detection_method":"D","context.country":"D","system.developer":"D","event.date_disclosed":"D","context.denominator_available":"D"},"field_citations":{"system.model_name":[{"source_index":0,"quote":"Matt Shumer, the former HyperWrite CEO turned AI investor, said on X that he was testing GPT-5.6 Sol in Ultra mode on July 10 when a subagent handling cleanup work ran rm -rf /Users/mattsdevbox."}],"event.agent_involved":[{"source_index":0,"quote":"Matt Shumer, the former HyperWrite CEO turned AI investor, said on X that he was testing GPT-5.6 Sol in Ultra mode on July 10 when a subagent handling cleanup work ran rm -rf /Users/mattsdevbox."}],"event.time_to_contain":[{"source_index":0,"quote":"He said he killed the process after one hour and 21 minutes."}],"event.incident_class":[{"source_index":0,"quote":"Matt Shumer, the former HyperWrite CEO turned AI investor, said on X that he was testing GPT-5.6 Sol in Ultra mode on July 10 when a subagent handling cleanup work ran rm -rf /Users/mattsdevbox."}],"impact.severity":[{"source_index":0,"quote":"Matt Shumer, the former HyperWrite CEO turned AI investor, said on X that he was testing GPT-5.6 Sol in Ultra mode on July 10 when a subagent handling cleanup work ran rm -rf /Users/mattsdevbox."}],"impact.severity_because":[{"source_index":0,"quote":"Matt Shumer, the former HyperWrite CEO turned AI investor, said on X that he was testing GPT-5.6 Sol in Ultra mode on July 10 when a subagent handling cleanup work ran rm -rf /Users/mattsdevbox."}],"event.summary":[{"source_index":0,"quote":"Matt Shumer, the former HyperWrite CEO turned AI investor, said on X that he was testing GPT-5.6 Sol in Ultra mode on July 10 when a subagent handling cleanup work ran rm -rf /Users/mattsdevbox."},{"source_index":0,"quote":"Then OpenAI's team asked him to stress test Ultra mode: the setting built for delegating heavier work to subagents that can operate with less direct oversight."},{"source_index":0,"quote":"He said he killed the process after one hour and 21 minutes."},{"source_index":0,"quote":"A large share of his files were already gone, and other AI agents helped him recover some of what had been deleted."},{"source_index":0,"quote":"The incident is worth treating carefully because the public record is thin."}]},"notes_on_gaps":"Single press source (Startup Fortune) relaying Matt Shumer's X post; the article itself states no independent report confirms the deletion claim, so all facts are his account. Not stated: model version beyond the Ultra mode label, tools granted, permissions, autonomy level, safeguards, lifecycle stage, date of disclosure, how the deletion was detected, trigger, root cause, responsible party, failure class, number of files or records lost, recovery extent, any OpenAI response or fix, and country. Judgements: deployment_context set to evaluation because the source describes a stress test requested by OpenAI; severity set to minor on the test-and-single-deployer reading, with no third party named; harm_class set to operational for lost files on one machine; B3 marked failed because the stop came after 81 minutes and was described as too late, though the source does not say the stop control itself failed. Failure class omitted because no source classes the act. Other incidents in the article (Hugging Face and Modal Labs breach, METR reward-hacking rate, UK AI Security Institute findings) were excluded as unrelated to this incident. [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 (1 report: Startup Fortune · subject to change)]"},"context":{"industry":null,"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":"went_wrong","text":"The subagent ran rm -rf against /Users/mattsdevbox, according to Shumer's X post.","quote":"Matt Shumer, the former HyperWrite CEO turned AI investor, said on X that he was testing GPT-5.6 Sol in Ultra mode on July 10 when a subagent handling cleanup work ran rm -rf /Users/mattsdevbox.","source_index":0,"grade":"C"},{"step":"done","text":"A large share of his files were already gone; other AI agents helped recover some of them.","quote":"A large share of his files were already gone, and other AI agents helped him recover some of what had been deleted.","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://startupfortune.com/matt-shumers-ai-agent-ran-rm-rf-and-deleted-his-macs-files-during-a-test/"]},"analysis":{"sections":{"timeline":[{"text":"Shumer was testing GPT-5.6 Sol in Ultra mode when a subagent ran rm -rf.","quote":"Matt Shumer, the former HyperWrite CEO turned AI investor, said on X that he was testing GPT-5.6 Sol in Ultra mode on July 10 when a subagent handling cleanup work ran rm -rf /Users/mattsdevbox.","source_index":0,"when":"2026-07-10"},{"text":"Shumer killed the process after one hour and 21 minutes.","quote":"He said he killed the process after one hour and 21 minutes.","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://startupfortune.com/matt-shumers-ai-agent-ran-rm-rf-and-deleted-his-macs-files-during-a-test/"]}}