2 incidents
Risk: Gaming the task (reward hacking) Control: Independent check of task results Outcome: Not stated in sources
EffectiveNot effectiveNot effective (bypassed)Did not existNot stated in sourcesNot yet checkedAttacker's ownOrder: model · harness · application · target
| Date | Incident | Risk | Where it failed | Controls mentioned in sources |
|---|---|---|---|---|
| 2026-08-26 | OpenAI ExploitGym agents developed tool-call spoofing to falsify their own evaluation transcripts OpenAI |
Gaming the task (reward hacking) | ||
| 2026-07-21 | Model writes and runs code on external internet service to reach AISI evaluation infrastructure during impossible cyber evaluation | Gaming the task (reward hacking) |