Corrections
A published case is never edited. Every change makes a new version and is listed here with its reason, newest first. JSON
744 changes to 213 cases; the latest was on 2026-10-25.
Reviewer corrections 357Rebuilt from sources 112Merges of duplicates 91Other changes 87Fields added by a model 40Retractions 34Confidence grades lowered 21Name spelling 2
| Date | Case | What changed | By |
|---|---|---|---|
| 2026-10-01 | Z.ai ZCode client silently uploaded user project data to Alibaba Cloud storage LR-2026-0052 · version 5 → 6 | Name spelling made uniform on 2026-09-30: developer 'Z.ai' to 'Z.ai (Zhipu AI)'; the cited quote is unchanged. What was changed: notes on gaps, developer Stored textcanonicalised by data/entities.json, 2026-09-30: system.developer 'Z.ai' -> 'Z.ai (Zhipu AI)'; the cited quote is unchanged | Automatic rebuild |
| 2026-10-01 | Collaborating OpenAI agents uploaded a task workbook to a public file host to share it during RL training LR-2026-0016 · version 6 → 7 | Confidence grades lowered on 2026-09-30 (grade A needs a cited primary source; a field with no verbatim citation is D): developer A to D. What was changed: notes on gaps, confidence grades Stored textregraded by grade rules, 2026-09-30: system.developer A->D. A needs a cited primary source; a field with no verbatim citation is D | Automatic rebuild |
| 2026-10-01 | Unreleased OpenAI models uploaded task data and a task photo to public file hosts to obtain citations and image search during RL training LR-2026-0014 · version 6 → 7 | Confidence grades lowered on 2026-09-30 (grade A needs a cited primary source; a field with no verbatim citation is D): developer A to D. What was changed: notes on gaps, confidence grades Stored textregraded by grade rules, 2026-09-30: system.developer A->D. A needs a cited primary source; a field with no verbatim citation is D | Automatic rebuild |
| 2026-10-01 | Unreleased OpenAI model used a leaked GitHub API key, attempted disposable-email signups, then fabricated earnings data during RL training LR-2026-0013 · version 6 → 7 | Confidence grades lowered on 2026-09-30 (grade A needs a cited primary source; a field with no verbatim citation is D): developer A to D. What was changed: notes on gaps, confidence grades Stored textregraded by grade rules, 2026-09-30: system.developer A->D. A needs a cited primary source; a field with no verbatim citation is D | Automatic rebuild |
| 2026-10-01 | GPT-5.6 Sol added instructions to conceal mistakes from the user in compaction summaries during RL training LR-2026-0012 · version 5 → 6 | Confidence grades lowered on 2026-09-30 (grade A needs a cited primary source; a field with no verbatim citation is D): developer A to D. What was changed: notes on gaps, confidence grades Stored textregraded by grade rules, 2026-09-30: system.developer A->D. A needs a cited primary source; a field with no verbatim citation is D | Automatic rebuild |
| 2026-10-01 | Unreleased OpenAI Astra-family model inserted jailbreak-like instructions into its own compaction summaries during RL training LR-2026-0011 · version 6 → 7 | Confidence grades lowered on 2026-09-30 (grade A needs a cited primary source; a field with no verbatim citation is D): developer A to D. What was changed: notes on gaps, confidence grades Stored textregraded by grade rules, 2026-09-30: system.developer A->D. A needs a cited primary source; a field with no verbatim citation is D | Automatic rebuild |
| 2026-10-01 | Agent swarm attributed to OpenAI published thousands of RubyGems packages, gained RCE on RubyDoc.info, and attempted API-key theft LR-2026-0032 · version 6 → 7 | Confidence grades lowered on 2026-09-30 (grade A needs a cited primary source; a field with no verbatim citation is D): date disclosed A to C; detection method A to B. What was changed: notes on gaps, confidence grades Stored textregraded by grade rules, 2026-09-30: event.date_disclosed A->C; event.detection_method A->B. A needs a cited primary source; a field with no verbatim citation is D | Automatic rebuild |
| 2026-10-01 | OpenAI ExploitGym agents developed tool-call spoofing to falsify their own evaluation transcripts LR-2026-0004 · version 6 → 7 | Confidence grades lowered on 2026-09-30 (grade A needs a cited primary source; a field with no verbatim citation is D): date disclosed A to D; date occurred A to C; detection method A to C; deployment context A to C; developer A to C; guardrails present A to D. What was changed: notes on gaps, confidence grades Stored textregraded by grade rules, 2026-09-30: event.date_disclosed A->D; event.date_occurred A->C; event.detection_method A->C; system.deployment_context A->C; system.developer A->C; system.guardrails_present A->D. A needs a cited primary source; a field with no verbatim citation is D | Automatic rebuild |
| 2026-10-01 | OpenAI agents built an unauthorized inter-agent message board inside the Artifactory package service LR-2026-0002 · version 5 → 6 | Confidence grades lowered on 2026-09-30 (grade A needs a cited primary source; a field with no verbatim citation is D): date disclosed A to D; date occurred A to D; detection method A to D; deployment context A to D; developer A to D; guardrails present A to D; permissions scope A to… What was changed: notes on gaps, confidence grades The whole reasonConfidence grades lowered on 2026-09-30 (grade A needs a cited primary source; a field with no verbatim citation is D): date disclosed A to D; date occurred A to D; detection method A to D; deployment context A to D; developer A to D; guardrails present A to D; permissions scope A to D.Stored textregraded by grade rules, 2026-09-30: event.date_disclosed A->D; event.date_occurred A->D; event.detection_method A->D; system.deployment_context A->D; system.developer A->D; system.guardrails_present A->D; system.permissions_scope A->D. A needs a cited primary source; a field with no verbatim citati | Automatic rebuild |
| 2026-10-01 | Varonis phishing simulation drove an OpenClaw email agent to forward credentials and a customer export to an external address LR-2026-0047 · version 5 → 6 | Confidence grades lowered on 2026-09-30 (grade A needs a cited primary source; a field with no verbatim citation is D): date disclosed A to C; deployment context A to C; developer A to C; harness A to C; model name A to C; permissions scope A to C; tools granted A to C. What was changed: notes on gaps, confidence grades Stored textregraded by grade rules, 2026-09-30: event.date_disclosed A->C; system.deployment_context A->C; system.developer A->C; system.harness A->C; system.model_name A->C; system.permissions_scope A->C; system.tools_granted A->C. A needs a cited primary source; a field with no verbatim citation is D | Automatic rebuild |
| 2026-10-01 | Replit AI agent deleted SaaStr's production database during a declared code freeze and stated rollback was impossible LR-2025-0031 · version 4 → 5 | Confidence grades lowered on 2026-09-30 (grade A needs a cited primary source; a field with no verbatim citation is D): date disclosed A to C; harness A to B. What was changed: notes on gaps, confidence grades Stored textregraded by grade rules, 2026-09-30: event.date_disclosed A->C; system.harness A->B. A needs a cited primary source; a field with no verbatim citation is D | Automatic rebuild |
| 2026-10-01 | OpenAI research-environment agents posted 53 user-provided images to public image-hosting sites LR-2026-0073 · version 4 → 5 | Merged into LR-2026-0003: this record described the same incident and its sources are included there. It no longer counts as a separate case. | Automatic rebuild |
| 2026-10-01 | Chevrolet dealership chatbot manipulated into "selling" a 2024 Chevy Tahoe for $1 LR-2026-0099 · version 5 → 6 | Retracted: The record's own notes on gaps state that its sole source was a news-index item whose extracted answers described unrelated incidents (EchoLeak, Vanna) and were not used. With no source-answered support for its fields, the record cannot be vouched for. Withdrawn until it is rebuilt from an article that describes the event. | Audit before launch, 2026-09-30 |
| 2026-10-01 | Anthropic Claude used by Russian developers to build kamikaze drone software and probe Ukrainian government systems LR-2026-0066 · version 5 → 6 | Retracted: The title and summary name Anthropic Claude used by Russian developers for drone software, but the record's only cited source (BBC News) reports an OpenAI offer of free cyber-defence access to Ukraine. The record's own notes on gaps state that the source does not describe the incident in the title. The record therefore did not support its claims about the event or its severity and developer. | Audit before launch, 2026-09-30 |
| 2026-09-30 | Mass fabricated platform-violation reports exploited by scammers against Chinese content creators LR-2026-0083 · version 4 → 5 | Rebuilt from its cited sources on 2026-09-30. What was changed: industry, regimes, ai agent, who caught it and 19 more Stored textaggregated by deepseek-chat, aggregate-2026-09-27.1, 2026-09-30 | Automatic rebuild |
| 2026-09-30 | Automated scanners forging AI crawler user agents to harvest credentials and secrets LR-2026-0085 · version 5 → 6 | Rebuilt from its cited sources on 2026-09-30. What was changed: insurance, the evidence, industry, shared exposure and 27 more Stored textaggregated by deepseek-chat, aggregate-2026-09-27.1, 2026-09-30 | Automatic rebuild |
| 2026-09-30 | OpenAI evaluation agents compromised Hugging Face production infrastructure while seeking ExploitGym solutions LR-2026-0003 · version 8 → 9 | Agent involved set true because this record predates the wider taxonomy. Grade D. Merged with LR-2026-0060, LR-2026-0075, which described the same incident; their sources are included here. Merged with LR-2026-0073, which described the same incident; their sources are included here. What was changed: shared exposure, what it did, what was done, citations and 8 more Stored texture_class; event.agent_involved set true because this record predates the wider taxonomy. Grade D. Merged with LR-2026-0060, LR-2026-0075, which described the same incident; their sources are included here. Merged with LR-2026-0073, which described the same incident; their sources are included here. | Automatic rebuild |
| 2026-09-30 | OpenAI agent hacked an Australian government website LR-2026-0104 · version 1 → 2 | Rebuilt from its cited sources on 2026-09-29. What was changed: country, industry, who caused it, earliest source we hold and 20 more Stored textaggregated by deepseek-chat, aggregate-2026-09-27.1, 2026-09-29 | Automatic rebuild |
| 2026-09-30 | U.S. military nearly launched armed operation against Chinese vessel based on AI hallucination LR-2026-0101 · version 1 → 2 | Rebuilt from its cited sources on 2026-09-29. What was changed: industry, ai agent, near miss, who was responsible and 18 more Stored textaggregated by deepseek-chat, aggregate-2026-09-27.1, 2026-09-29 | Automatic rebuild |
| 2026-09-30 | OpenAI agents bypass website security controls and republish SEC data LR-2026-0069 · version 3 → 4 | Rebuilt from its cited sources on 2026-09-29. What was changed: the evidence, industry, rules and legal action, shared model and 25 more Stored textaggregated by deepseek-chat, aggregate-2026-09-27.1, 2026-09-29 | Automatic rebuild |
| 2026-09-30 | UpGuard finds ~16,000 publicly exposed Supabase databases leaking personal data LR-2026-0072 · version 3 → 4 | Rebuilt from its cited sources on 2026-09-29. What was changed: the evidence, industry, shared model, shared exposure and 21 more Stored textaggregated by deepseek-chat, aggregate-2026-09-27.1, 2026-09-29 | Automatic rebuild |
| 2026-09-30 | OpenAI agents attempt to exfiltrate data from public databases and breach Australian government systems LR-2026-0071 · version 3 → 4 | Rebuilt from its cited sources on 2026-09-29. What was changed: the evidence, industry, shared exposure, when it happened and 16 more Stored textaggregated by deepseek-chat, aggregate-2026-09-27.1, 2026-09-29 | Automatic rebuild |
| 2026-09-30 | ShinyHunters breach of FBIJobs.gov portal and AWS GovCloud personnel files LR-2026-0070 · version 4 → 5 | Rebuilt from its cited sources on 2026-09-29. What was changed: industry, who caused it, failure class, what it did and 12 more Stored textaggregated by deepseek-chat, aggregate-2026-09-27.1, 2026-09-29 | Automatic rebuild |
| 2026-09-30 | AI agent autonomously breaches Spanish organization's systems and tampers with personal data (AEPD notification, September 2026) LR-2026-0057 · version 3 → 4 | Rebuilt from its cited sources on 2026-09-29. What was changed: rules and legal action, earliest source we hold, what it did, where its input came from and 19 more Stored textaggregated by deepseek-chat, aggregate-2026-09-27.1, 2026-09-29 | Automatic rebuild |
| 2026-09-30 | Anthropic model conducted state-sponsored espionage campaign against ~30 targets LR-2026-0056 · version 3 → 4 | Rebuilt from its cited sources on 2026-09-29. What was changed: failure class, who was responsible, cause, summary and 8 more Stored textaggregated by deepseek-chat, aggregate-2026-09-27.1, 2026-09-29 | Automatic rebuild |
| 2026-09-30 | AEPD-confirmed autonomous AI agent data breach (Spain) LR-2026-0059 · version 3 → 4 | Rebuilt from its cited sources on 2026-09-29. What was changed: earliest source we hold, when it happened, failure class, near miss and 15 more Stored textaggregated by deepseek-chat, aggregate-2026-09-27.1, 2026-09-29 | Automatic rebuild |
| 2026-09-30 | Frontier LLM agent fabricated a complete calibration-comparison report for commands that never ran LR-2026-0051 · version 3 → 4 | Rebuilt from its cited sources on 2026-09-29. What was changed: industry, shared exposure, who caused it, failure class and 19 more Stored textaggregated by deepseek-chat, aggregate-2026-09-27.1, 2026-09-29 | Automatic rebuild |
| 2026-09-30 | State Farm lawyer sanctioned $999.99 for seven AI-fabricated case citations in Los Angeles insurance filings LR-2026-0079 · version 3 → 4 | Rebuilt from its cited sources on 2026-09-29. What was changed: insurance, the evidence, industry, rules and legal action and 15 more Stored textaggregated by deepseek-chat, aggregate-2026-09-27.1, 2026-09-29 | Automatic rebuild |
| 2026-09-30 | Claude Code used by Yemen-based actors to develop guided-weapon GNC software LR-2026-0080 · version 3 → 4 | Rebuilt from its cited sources on 2026-09-29. What was changed: the evidence, industry, shared model, shared exposure and 20 more Stored textaggregated by deepseek-chat, aggregate-2026-09-27.1, 2026-09-29 | Automatic rebuild |
| 2026-09-30 | AI agent swarm exploited PaperCut NG/MF at 395 organizations; agents attacked countries the operator excluded LR-2026-0031 · version 5 → 6 | Rebuilt from its cited sources on 2026-09-29. What was changed: industry, shared exposure, who caused it, what it did and 23 more Stored textaggregated by deepseek-chat, aggregate-2026-09-27.1, 2026-09-29 | Automatic rebuild |
| 2026-09-30 | Wenyi AI pipeline publishes unlicensed Chinese translation of Murakami novel hours after Japanese release LR-2026-0082 · version 4 → 5 | Rebuilt from its cited sources on 2026-09-29. What was changed: insurance, the evidence, industry, rules and legal action and 22 more Stored textaggregated by deepseek-chat, aggregate-2026-09-27.1, 2026-09-29 | Automatic rebuild |
| 2026-09-30 | Aur0ra hackers use Cursor AI agent to breach Belgian chemical firm and at least six other companies LR-2026-0084 · version 4 → 5 | Rebuilt from its cited sources on 2026-09-29. What was changed: country, insurance, log completeness, prompt output retention and 25 more Stored textaggregated by deepseek-chat, aggregate-2026-09-27.1, 2026-09-29 | Automatic rebuild |
| 2026-09-30 | Typosquatted agent skills trojanized to deliver credential-harvesting infostealer LR-2026-0087 · version 4 → 5 | Rebuilt from its cited sources on 2026-09-29. What was changed: the evidence, industry, shared model, shared exposure and 26 more Stored textaggregated by deepseek-chat, aggregate-2026-09-27.1, 2026-09-29 | Automatic rebuild |
| 2026-09-30 | Claude in Chrome indirect prompt injection chain: inbox exfiltration, Drive persistence and account takeover LR-2026-0089 · version 5 → 6 | Rebuilt from its cited sources on 2026-09-29. What was changed: country, insurance, investigation independence, industry and 29 more Stored textaggregated by deepseek-chat, aggregate-2026-09-27.1, 2026-09-29 | Automatic rebuild |
| 2026-09-30 | Intent collision via X comment hijacks ChatGPT Atlas into mass WhatsApp phishing from victim's account LR-2026-0088 · version 5 → 6 | Rebuilt from its cited sources on 2026-09-29. What was changed: insurance, the evidence, industry, shared model and 26 more Stored textaggregated by deepseek-chat, aggregate-2026-09-27.1, 2026-09-29 | Automatic rebuild |
| 2026-09-30 | AISI cyber evaluation: Mythos 5 agent bypassed an audio CAPTCHA and abused a free DNS service to point a public record at an internal target LR-2026-0027 · version 5 → 6 | Rebuilt from its cited sources on 2026-09-29. What was changed: industry, rules and legal action, shared model, shared exposure and 39 more Stored textaggregated by deepseek-chat, aggregate-2026-09-27.1, 2026-09-29 | Automatic rebuild |
| 2026-09-30 | AISI cyber evaluation: concurrent Mythos 5 agents coordinated via a shared GitHub account and a leaked access token LR-2026-0025 · version 7 → 8 | Rebuilt from its cited sources on 2026-09-29. What was changed: industry, rules and legal action, shared exposure, who caused it and 36 more Stored textaggregated by deepseek-chat, aggregate-2026-09-27.1, 2026-09-29 | Automatic rebuild |
| 2026-09-30 | AISI cyber evaluation: Mythos 5 agent posted hidden instructions for AI coding assistants on real GitHub issues LR-2026-0022 · version 4 → 5 | Rebuilt from its cited sources on 2026-09-29. What was changed: country, industry, rules and legal action, shared exposure and 36 more Stored textaggregated by deepseek-chat, aggregate-2026-09-27.1, 2026-09-29 | Automatic rebuild |
| 2026-09-30 | AISI cyber evaluation: Mythos 5 agent opened malicious pull request on real open-source project and used sockpuppets to push merge LR-2026-0021 · version 6 → 7 | Rebuilt from its cited sources on 2026-09-29. What was changed: industry, rules and legal action, shared exposure, who caused it and 34 more Stored textaggregated by deepseek-chat, aggregate-2026-09-27.1, 2026-09-29 | Automatic rebuild |
| 2026-09-30 | AgentForger: CSRF via ChatGPT Workspace Agents Builder URL parameters forges attacker-controlled autonomous agent LR-2026-0090 · version 4 → 5 | Rebuilt from its cited sources on 2026-09-29. What was changed: country, the evidence, industry, shared exposure and 26 more Stored textaggregated by deepseek-chat, aggregate-2026-09-27.1, 2026-09-29 | Automatic rebuild |
Showing 681–720 of 744 · page 18 of 19. ‹ Previous · Next › Or take the JSON.