Search
On September 17, Dario Amodei, CEO of Anthropic, proposed permanently embedding independent third-party safety assessors within leading AI companies, granting them access akin to internal risk teams and allowing them to autonomously publish findings with limited redactions. However, multiple regulatory and audit experts pointed out that such assessors lack the enforcement powers held by banking regulators, unable to halt model training or releases. Moreover, the practice of allowing companies to select assessment agencies, control their scope of visibility, and freely disregard conclusions could lead to conflicts of interest, resulting in 'audit whitewashing.' Experts argue that AI regulation requires establishing independent qualification standards, review criteria, and accountability mechanisms to ensure the objectivity and effectiveness of assessments.
-
C114 Communication Network -
Communication Home
