The Government Accountability Office (GAO) has issued its artificial intelligence (AI) accountability framework for Federal agencies and other entities to ensure responsible, equitable, traceable, reliable, and governable AI.
The AI accountability framework is organized around four components that address governance, data, performance, and monitoring. For each of these principles, GAO described practices for Federal agencies and other entities that are considering, selecting, and implementing AI.
“For each principle, the framework includes the following: key practices, which we developed by synthesizing information and identifying at least two sources that noted the importance of a certain practice in implementing AI systems; key questions, which we developed from information provided during the [Comptroller General of the United States] forum, interviews with experts, and documents; and audit procedures, which we developed by reviewing the types of evidence noted in the Government Auditing Standards,” wrote GAO.
According to GAO, AI systems can “pose unique challenges for independent assessments and audits to promote accountability because their inputs and operations are not visible to the user.” The lack of transparency can limit the ability of auditors to detect errors or misuse. Additionally, AI has the ability to amplify existing biases and concerns related to civil liberties, ethics, and social disparities.
“GAO’s objective was to identify key practices to help ensure accountability and responsible AI use by Federal agencies and other entities involved in the design, development, deployment, and continuous monitoring of AI systems,” wrote GAO.
The work for the framework was conducted from February 2020 to June 2021, in accordance with all sections of GAO’s Quality Assurance Framework which requires that GAO plan and perform engagement to obtain sufficient and appropriate evidence to meet stated objectives and discuss any limitations in the work.