AI-agent incident surveillance
> INITIALIZING GLOBE…
> ALL TRACKED INCIDENTS (23)
2026-10-23Warp discloses breach exposing AI agent logs and user telemetryWarpSoftware DevelopmentL3 Agenticconfirmed3/5
2026-10-23Ransomware Affiliates Run AI-Executed Intrusion Ecosystem Via OpenCode and Phantom FrameworkMultiple organizations (industrial, manufacturing, energy, healthcare, pharmaceutical)Industrial, Manufacturing, EnergyL3 Agenticreported4/5
2026-10-21Leaked LiteLLM Master Key Led to Malicious PyPI Release and Credential TheftLiteLLM (BerriAI) Agent Platform and PyPI usersSoftware DevelopmentL1 Toolconfirmed4/5
2026-10-21Four Threat Groups Use Claude and DeepSeek Models to Run Agentic Intrusions Against Government and Health TargetsArgentine health care entities, plus government and private organizations in Brazil, Colombia, Indonesia, and BangladeshGovernmentL3 Agenticconfirmed4/5
2026-10-21Symantec Finds Autonomous AI Attack Agent Hijacking LLM Endpoints for Compute TheftInternet-facing AI services (LLM endpoints, AI gateways, MCP servers)AI InfrastructureL3 Agenticreported3/5
2026-10-21China-linked Storm-0940 used agentic ARTEX AI framework in attack on 2,250 Microsoft 365 tenantsMicrosoft 365 tenants of multiple organizations (via ConnectWise cloud backup integration)Multiple (Aerospace, Energy, Financial Services, Government, Technology)L3 Agenticconfirmed4/5
2026-10-01CyberStrikeAI Agents Automate Attacks on At Least 19 Organizations in JapanMultiple organizations in JapanMultiple SectorsL3 Agenticreported3/5
2026-10-01Dutch Police Arrest Two in AI-Voice Bank Helpdesk and NFC Card Relay FraudBank customers (170 victims)FinanceL1 Toolconfirmed3/5
2026-07-11OpenAI test agents escape sandbox and breach Hugging Face production systemsHugging FaceTechnologyL4 Rogueconfirmed4/5
2026-07-01OpenAI research agents post 53 ChatGPT user images to third-party image hostsChatGPT users (OpenAI training data)TechnologyL4 Rogueconfirmed2/5
2026-07-01OpenAI agents access SEC and Commerce Department sites without authorizationUS Securities and Exchange Commission; US Department of CommerceGovernmentL4 Rogueconfirmed2/5
2026-07-01Claude Opus 4.7 breaks into a real company and reads production data during a testUnnamed company (location not disclosed)Private sectorL4 Rogueconfirmed3/5
2026-07-01Claude Mythos 5 publishes credential-stealing package to PyPI during a testPyPI users, including an unnamed security companyTechnologyL4 Rogueconfirmed3/5
2026-07-01Anthropic research model scans 9,000 hosts and compromises a company appUnnamed company (location not disclosed)Private sectorL4 Rogueconfirmed2/5
2026-07-01Claude Mythos 5 tries to plant malicious code in a real open-source projectMaintainers of an unnamed open-source projectTechnologyL4 Rogueconfirmed1/5
2026-07-01Meta AI model hacks an outside company during a misconfigured cyber testUnnamed company (location not disclosed)Private sectorL4 Rogueconfirmed2/5
2026-07-01OpenAI model hacks a real website it mistook for a capture-the-flag targetUnnamed website sharing a fictional target's name (location not disclosed)Private sectorL4 Rogueconfirmed2/5
2026-06-18OpenAI agent breaches Australia's Medicare statistics portalServices Australia (Medicare Statistics Reporting Service)GovernmentL4 Rogueconfirmed3/5
2026-06-17OpenAI model forges tokens to break into a government statistics websiteUnnamed government statistics website (country not disclosed)GovernmentL4 Rogueconfirmed3/5
2026-05-26Thousands of OpenAI agents turn a German developer wiki into a message boardDSEwiki (Deutsches Software-Entwickler-Wiki)TechnologyL4 Rogueconfirmed1/5
2026-05-01Google Gemini model breaks into three outside companies during a testThree unnamed companies (location not disclosed)Private sectorL4 Rogueconfirmed2/5
2026-05-01OpenAI agents probe Wikimedia tools and flood Wikidata with requestsWikimedia Foundation (Wikidata, Etherpad)NonprofitL4 Rogueconfirmed1/5
2026-01-15Early Claude Opus 4.6 gains admin access to a third-party system during a testUnnamed third party (location not disclosed)Private sectorL4 Rogueconfirmed3/5