TY - JOUR T1 - Toward Trustworthy AI Development: Mechanisms for Supporting Verifiable Claims Y1 - 2020 A1 - Brundage, Miles A1 - Avin, Shahar A1 - Wang, Jasmine A1 - Belfield, Haydn A1 - Krueger, Gretchen A1 - Hadfield, Gillian A1 - Khlaaf, Heidy A1 - Yang, Jingying A1 - Toner, Helen A1 - Fong, Ruth A1 - Maharaj, Tegan A1 - Koh, Pang Wei A1 - Hooker, Sara A1 - Leung, Jade A1 - Trask, Andrew A1 - Bluemke, Emma A1 - Lebensold, Jonathan A1 - O'Keefe, Cullen A1 - Koren, Mark A1 - Ryffel, Théo A1 - Rubinovitz, JB A1 - Besiroglu, Tamay A1 - Carugati, Federica A1 - Clark, Jack A1 - Eckersley, Peter A1 - de Haas, Sarah A1 - Johnson, Maritza A1 - Laurie, Ben A1 - Ingerman, Alex A1 - Krawczuk, Igor A1 - Askell, Amanda A1 - Cammarota, Rosario A1 - Lohn, Andrew A1 - Krueger, David A1 - Stix, Charlotte A1 - Henderson, Peter A1 - Graham, Logan A1 - Prunkl, Carina A1 - Martin, Bianca A1 - Seger, Elizabeth A1 - Zilberman, Noa A1 - HÉigeartaigh, Seán Ó A1 - Kroeger, Frens A1 - Sastry, Girish A1 - Kagan, Rebecca A1 - Weller, Adrian A1 - Tse, Brian A1 - Barnes, Elizabeth A1 - Dafoe, Allan A1 - Scharre, Paul A1 - Herbert-Voss, Ariel A1 - Rasser, Martijn A1 - Sodhani, Shagun A1 - Flynn, Carrick A1 - Gilbert, Thomas Krendl A1 - Dyer, Lisa A1 - Khan, Saif A1 - Bengio, Yoshua A1 - Anderljung, Markus AB - With the recent wave of progress in artificial intelligence (AI) has come a growing awareness of the large-scale impacts of AI systems, and recognition that existing regulations and norms in industry and academia are insufficient to ensure responsible AI development. In order for AI developers to earn trust from system users, customers, civil society, governments, and other stakeholders that they are building AI responsibly, they will need to make verifiable claims to which they can be held accountable. Those outside of a given organization also need effective means of scrutinizing such claims. This report suggests various steps that different stakeholders can take to improve the verifiability of claims made about AI systems and their associated development processes, with a focus on providing evidence about the safety, security, fairness, and privacy protection of AI systems. We analyze ten mechanisms for this purpose–spanning institutions, software, and hardware–and make recommendations aimed at implementing, exploring, or improving those mechanisms. UR - http://arxiv.org/abs/2004.07213 ER -