Facebook’s Civil Rights Audit universities to join the Align consortium, including: Georgia Tech, University of Illinois at Urbana–Champaign, and Columbia. In addition to focusing on increasing diversity overall in AI, Facebook states that it has also increased hiring from civil society including nonprofits, research, and advocacy organizations that work closely with major civil rights institutions on emerging technology-related challenges — and these employees are actively engaged in the Responsible AI organization. B. Auditor Observations It is important that Facebook has publicly acknowledged that AI can be biased and discriminatory and that deploying AI and machine learning models brings with it a responsibility to ensure fairness and accountability. The Auditors are encouraged that Facebook is devoting resources to studying responsible AI methodologies and engaging with external experts regarding best practices. When it comes to Facebook’s own algorithms and machine learning models, the Auditors cannot speak to the effectiveness of any of the pilots Facebook has launched to better identify and address potential sources of bias or discriminatory outcomes. (Both because the pilots are still in nascent stages and the Auditors have not had full access to the full details of these programs.) The Auditors do, however, credit Facebook for taking steps to explore ways to improve Facebook’s AI infrastructure and develop processes designed to help spot and correct biases, skews, and inaccuracies in Facebook’s models. That being said, the Auditors strongly believe that processes and guidance designed to prompt issue-spotting and help resolve fairness concerns must be mandatory (not voluntary) and company-wide. That is, all teams building models should be required to follow comprehensive best practice guidance and existing algorithms and machinelearning models should be regularly tested. This includes both guidance in building models and systems for testing models. And while the Auditors believe it is important for Facebook to have a team dedicated to working on AI fairness and bias issues, ensuring fairness and non-discrimination should also be a responsibility for all teams. To that end, the Auditors recommend that training focused on understanding and mitigating against sources of bias and discrimination in AI should be mandatory for all teams building algorithms and machine-learning models at Facebook and part of Facebook’s initial onboarding process. Landing on a set of widely accepted best practices for identifying and correcting bias or discrimination in models or for handling sensitive data questions is likely to take some time. Facebook can and should be a leader in this space. Moreover, Facebook cannot wait for consensus (that may never come) before building an internal infrastructure to ensure that the algorithms and machine learning models it builds meet minimum standards already known to help avoid bias pitfalls (e.g., use of inclusive data sets, critical assessment of model assumptions and inferences for potential bias, etc.). Facebook has an existing responsibility to ensure that the algorithms and machine learning models that can have important impacts on billions of people do not have unfair or adverse consequences. The Auditors think Facebook needs to approach these issues with a greater sense of urgency. There are steps it can take 81