Toward Trustworthy AI Development: Mechanisms for Supporting Verifiable Claims

With the recent wave of progress in artificial intelligence (AI) has come a awareness of the large-scale impacts of AI systems, and recognition existing regulations and norms in industry and academia are insufficient ensure responsible AI development. In order for AI developers to earn trust system users, customers, civil society, governments, and other that they are building AI responsibly, they will need to make claims to which they can be held accountable. Those outside of a organization also need effective means of scrutinizing such claims. This suggests various steps that different stakeholders can take to improve verifiability of claims made about AI systems and their associated processes, with a focus on providing evidence about the safety,, fairness, and privacy protection of AI systems. We analyze ten for this purpose--spanning institutions, software, and hardware--and recommendations aimed at implementing, exploring, or improving those.