Improving verifiability in AI development (openai.com)

Multi-stakeholder report outlining 10 mechanisms to improve verifiability of AI safety, security, fairness, and privacy claims for use by developers, policymakers, and civil society in evaluating AI systems.