Incidents Models AI Safety AI Risk Trends Newsletter Regulatory Updates Raw Data Method Corrections

Cases · Stage of use: in testing · How the AI agent failed: obeyed a harmful order · Added to the register (date we first published it): last 7 days

All cases ›Stage of use: in testing✕›How the AI agent failed: obeyed a harmful order✕›Added to the register (date we first published it): last 7 days✕ 0 cases in this selection

Nothing matches. Start again.

Filter further

How the AI agent failed pinned

escaped its sandbox3
lied about or hid what it did1

Added to the register (date we first published it) pinned

last 30 days2

Get the daily brief by email

Every quoted sentence links to its stored source · schema v0.9.0 · JSON