AI Safety IndicesAI Safety LevelPreparedness Framework Savotageassets.anthropic.comhttps://assets.anthropic.com/m/377027d5b36ac1eb/original/Sabotage-Evaluations-for-Frontier-Models.pdf
