[Anthropic CEO Proposes 'Bank-Like' AI Regulation, Experts Question Evaluators' Lack of Enforcement Power] Deep Tide TechFlow reports that on September 17, Anthropic CEO Dario Amodei proposed embedding independent third-party safety evaluators long-term within cutting-edge AI companies, granting them access to internal systems and allowing them to independently publish investigation results. However, multiple regulatory and auditing experts pointed out that such evaluators lack the legal authority to halt model training or deployment, fundamentally differing from the administrative enforcement powers of banking regulators. Additionally, allowing companies to select evaluation agencies, define the scope of reviews, and control subsequent handling could lead to conflicts of interest and "audit whitewashing." Experts believe that AI regulation still requires the establishment of independent accreditation, review standards, and accountability mechanisms.