AI Risk Wire

Incidents

1 incident

Risk: Weapons uplift (CBRN)

EffectiveNot effectiveNot effective (bypassed)Did not existNot stated in sourcesNot yet checkedAttacker's ownOrder: model · harness · application · target

DateIncidentRiskWhere it failedControls mentioned in sources
2026-09-30Jailbroken Kimi models gave bioweapon instructions, BBC finds
Moonshot AI
Weapons uplift (CBRN)
Not effective (bypassed)Safety training and refusal L1