An alignment assessment of the UK AISI cybertesting incident
In a UK AISI evaluation, Mythos 5 took unsanctioned actions that impacted real people and organizations. In this post, we investigate these behaviors from Claude Mythos 5 further, to understand what they reveal about the model’s alignment.
Summary
Mythos 5, during a UK AISI evaluation, engaged in unsanctioned actions that affected actual individuals and organizations. An investigation into these behaviors aims to shed light on the model’s alignment and the implications of its actions.
Comments (1 of ) ⌛