Attention: You are using an outdated browser, device or you do not have the latest version of JavaScript downloaded and so this website may not work as expected. Please download the latest software or switch device to avoid further issues.
| 1 Apr 2023 | |
| AI |
Overview
Use of artificial intelligence (AI) has the potential to reduce cost, ease compliance burdens, and improve effectiveness of operations and service delivery – garnering much enthusiasm and expectation for federal agencies to incorporate AI across government. High-quality, reliable, and accessible data feed these technologies. Strong data governance and management practices can address ethics and equity concerns by providing a way for the government to detect bias in AI algorithms or discriminatory practices within agency programs. In order for agencies to be prepared to address the mounting pressure to realize the benefits of AI, they must first have a wealth of high-quality data, governed by effective principles, to develop AI technologies and tools.
Both the current and previous administrations have signaled increased interest in strengthening the use of AI in the U.S. government, and many agencies are taking steps to leverage AI technologies to improve internal processes and program operations. Executive Orders 13859, “Maintaining American Leadership in Artificial Intelligence,” in 2019 and 13960, “Promoting the Use of Trustworthy Artificial Intelligence in the Federal Government,” in 2020 from the last administration require federal agencies to publish publicly-available AI strategies and develop AI use case inventories, respectively. Further, in October 2022, the current White House released Blueprint for the AI Bill of Rights and announced steps for agency action to ensure AI technologies are held accountable and do not cause harm or perpetuate harmful biases. These efforts, in concert with the executive orders to promote transparency and thoughtful development of AI, provide a foundation for responsible implementation of AI across the federal government.
Many agencies have complied with these requirements, developing various public plans to incorporate AI into operations. Certain agencies demonstrate more mature AI governance frameworks with clearly aligned data management strategies and AI plans, centralized AI offices, oversight processes, and established technical capacity. Others are still developing an approach to AI, with strategies that primarily outline general goals and do not address implementation plans. Whether broad strategies or detailed implementation plans, the steps to organize and orient agencies’ data as it fits with overall AI goals are critical to ensure agencies have a data-focused foundation on which to build their AI capabilities.
These government-wide guiding documents, agency strategies, and AI use case inventories lay the groundwork for transparency and accountability, but specific details and progress updates are still unavailable to the public. According to one comprehensive review, there is insufficient publicly-available technical documentation
of the agencies’ methods for AI – and the unclassified data assets that are used to train AI technology are not effectively disclosed to the public. This affects public understanding of AI and ultimately hinders the ability of the U.S. government to achieve its goal of trustworthy AI use.
Beyond transparency, there are many more challenges that agencies face to adopt AI. A report from the Brookings Institution’s Artificial Intelligence and Emerging Technology Initiative describes a lack of technical standards, data limitations, cultural barriers to organizational change, underdeveloped responsible AI principles, procurement obstacles, and insufficiently trained employees.
To reflect existing practices that address these challenges, this resource is broken down by three key components of safe, effective adoption of AI: high-quality data, effective governance principles, and technical capacity. Each category then provides an example of an agency that has developed and implemented a relevant approach; agencies highlighted for successful strategies were selected based on the availability and level of detail in public-facing AI strategies and use cases.
