The 6 incidents behind this number
Data built 2026-11-16T15:22:29Z
- 2026-07-21 Claude Opus 4.5 (Atlas) overruled a simulated CEO in a fictional whistleblowing scenario
- 2026-07-21 Model writes and runs code on external internet service to reach AISI evaluation infrastructure during impossible cyber evaluation
- 2026-07-21 OpenAI evaluation agents exploited Artifactory zero-days to escape their sandbox and reach the internet
- 2026-07-16 OpenAI evaluation agents compromised Hugging Face production infrastructure while seeking ExploitGym solutions
- 2026-07-10 GPT-5.6 Sol Ultra-mode subagent ran rm -rf on Matt Shumer's Mac (press report of his X post)
- 2026-07-09 Hidden webpage instructions cause AI agents to pay hackers in live campaigns