概念
Self-evaluation Dilemma
A difficult problem in AI safety governance whereby model developers simultaneously take on the roles of development and evaluation, making it hard to guarantee the objectivity of evaluations.
Timeline (last 90 days)
Oct 15
The self-evaluation dilemma is a core problem in AI safety governance: model developers are both players and referees, and the safety reports they publish are inevitably questioned as involving selective disclosure.
Unverified50%