Dario Amodei, CEO of Anthropic, has proposed giving independent third-party evaluators employee-level access to the company’s offices, laptops, and AI development environments. In an essay published on September 12, 2026 titled « We Must Pace the Frontier, » Anthropic committed to embedding evaluators with ongoing access rather than conducting periodic audits. The evaluators would be able to publish their findings with minimal redaction and without editorial control from Anthropic. METR, an AI safety evaluation organization, was named as a potential evaluator under this framework. This shift from episodic to continuous evaluation aims to create verifiable safety commitments, though questions remain about implementation boundaries and publication rights.
Source: Read the original article
