While some AI harms are already widely known, such as deepfakes, scams and biased results, the report said that “as general-purpose AI becomes more capable, evidence of additional risks is ...