Anthropic sabotage evaluations find current controls largely contain tested attacks

DoomBench assesses “Anthropic sabotage evaluations find current controls largely contain tested attacks” as evidence moving away from doom, with magnitude 30 and confidence 90 out of 100 in the safety and alignment category.

The DoomBench assessment of “Anthropic sabotage evaluations find current controls largely contain tested attacks” is based on reporting from Anthropic and records the editorial rationale, source quality, attribution, and revision history.

DoomBench summarizes “Anthropic sabotage evaluations find current controls largely contain tested attacks” as follows: Anthropic demonstrated evaluations for decision sabotage, code sabotage, sandbagging, and oversight manipulation...