Miles Brundage Questions 'Most Aligned' AI Claims

Miles Brundage posted that uncertainty about eval awareness, monitorability and related issues means companies should not describe their systems as most aligned. He noted both OpenAI and Anthropic have made similar statements recently. Brundage is identified in the post as an AI policy researcher and Executive Director of AVERI, a nonprofit focused on third-party auditing and evaluation of frontier AI systems. The statement is contained in a single public post on X.