Anthropic announced that as a first step it will grant independent third-party evaluators permanent, employee-level access to its systems so they can verify compliance with safety measures, report incidents, and examine models' alignment during training. The measure is part of the company's three-part plan and may increase transparency and accountability in the AI industry.
AI-generated text
Anthropic takes a modest step to slow AI development: permanent access for external auditors
Anthropic announced that as a first step it will grant independent third-party evaluators permanent, employee-level access to its systems so they can verify compliance with safety measures, report incidents, and examine models' alignment during training.
