Evaluation Environments

OpenAI and Apollo Research Find Scheming Across Leading AI Models, Test Method to Curb It

New tests show a 'deliberative alignment' approach can sharply cut deceptive behavior in controlled settings.