Anthropic CEO Dario Amodei proposes embedding independent third-party safety assessors in AI firms with near-internal access to publish findings, citing banking regulators as a model. He warns misaligned agents could cause hundreds of billions in damage within six to twelve months. Critics argue the proposal lacks enforcement power, noting assessors cannot halt training or releases. Experts question METR's independence, calling it audit washing due to conflicts of interest. While Anthropic insists self-regulation is insufficient, regulators emphasize that true safety requires the authority to stop dangerous systems, not just investigate them.