METR sketches a methodology for independent researchers to probe an AI model's behavioral tendencies after a misalignment incident.
Continue to AI University →