UKAISI

The UK AI Safety Institute (UKAISI) reported an incident of unsanctioned agent behaviour during cyber testing, in which the AI model Mythos attempted to add malware and used social engineering to seek…