VIENNA AGENTIC INCIDENTS DATABASE
// public database of incidents involving autonomous AI agents, scored on the VAID scale 0–8
TOTAL RECORDS: 30  |  SHOWING: 8  |  distribution →
[ RESET ]
REFVAIDDATETITLEDOMAINSTATUS
VAID-2026-0035 1 2026-07-25 AISI evaluation agents take unsanctioned actions against real people and projects
DECENGSCBUNA Anthropic Mythos 5 and OpenAI GPT-5.6 Sol in specially configured evaluations · United Kingdom
 research under investigation
VAID-2026-0033 3 2026-07-09 OpenAI evaluation agents coordinate an intrusion into Hugging Face
DECDEXENGLOCSCBUCAUNA Multiple OpenAI models in an internal ExploitGym evaluation deployment
 research under investigation
VAID-2026-0036 3 2026-04 (day unknown) Anthropic identifies three real-world intrusions during cyber evaluations
DEXLOCUNA Claude Opus 4.7, Mythos 5 and an internal research test model
 research under investigation
VAID-2025-0039 2 Not established ROME training agent opens an external tunnel and diverts GPUs to cryptocurrency mining
SCBUNA ROME agentic model in reinforcement-learning training
 research reported
VAID-2025-0018 0 Not established Anthropic documents agentic misalignment in simulated companies
DECUNA 16 models from several developers, including Anthropic, OpenAI, Google, Meta and xAI · research environment
 research reported
VAID-2025-0016 0 Not established Reasoning models interfere with shutdown scripts in Palisade tests
LOCSCB OpenAI o3, o4-mini and codex-mini, with other models tested for comparison · research environment
 research reported
VAID-2024-0013 0 Not established Apollo Research documents scheming in controlled agent evaluations
DECSCB o1, Claude 3.5 Sonnet, Claude 3 Opus, Gemini 1.5 Pro and Llama 3.1 405B · research environment
 research reported
VAID-2024-0009 2 Not established AI Scientist experiments spawn uncontrolled processes and attempt runtime extensions
SCBSRP The AI Scientist, developed by researchers affiliated with Sakana AI, Oxford and UBC · Tokyo, JP
 research reported