Artificial Intelligence (AI) algorithms are being deployed in executive branch agencies at a brisk pace and with no executive branch account for their use. The proliferation of AI in government raises civil rights concerns because it has been found—at a general level—to succumb to racial and gender biases when AI algorithms are incompetently or intentionally trained. Policy solutions have been put forth to mitigate the issue of such AI uses in government, some of which are in the process of being implemented. Despite these gains, the political branches of the federal government have limited time to act before their primary policymaking authority is diluted or potentially ceded to the Judiciary. This essay argues that in the absence of timely action of the Executive or Legislative branches to establish procedures to mitigate administrative agency AI accountability and transparency gaps, the Judiciary may dictate such procedures via remands under the administrative record provision of the Administrative Procedure Act of 1946 as it first did 50 years ago with informal adjudications in Citizens to Preserve Overton Park v. Volpe, 401 U.S. 402 (1971).
GW Paper Series
97 Notre Dame Law Review Reflection (forthcoming 2022)