Evaluations of new AI models’ capabilities and risks are an important cornerstone of safety-focused AI policy. Currently, their future as part of the policy platform faces peril for four reasons: Entanglements with a broad AI safety ecosystem, structural incentives favouring less helpful evals, susc

Evaluations of new AI models’ capabilities and risks are an important cornerstone of safety-focused AI policy. Currently, their future as part of the policy platform faces peril for four reasons: Entanglements with a broad AI safety ecosystem, structural incentives favouring less helpful evals, susc

Anton Leicht suggests that a more defensive approach to the first episode of frontier AI policy might have been more prudent.

Anton Leicht suggests that a more defensive approach to the first episode of frontier AI policy might have been more prudent.