A Federal Oversight Report Warns of Gaps in the U.S. Government's AI Policies
A recently released report by the Government Accountability Office (GAO) has highlighted significant gaps in the U.S. government's policies on responsibly acquiring and utilising artificial intelligence (AI) from the private sector. Despite plans for a vast expansion of reliance on AI, the lack of a government-wide standard for AI purchases poses potential risks to American security, according to the GAO, the government's primary accountability watchdog.
The 96-page report catalogued over 200 existing non-military AI and machine learning uses across nearly two dozen agencies, along with more than 500 planned applications. The report emphasised the urgent need for consistent policies, citing potential threats to national security, worker displacement, election misinformation, and algorithmic biases affecting vulnerable populations.
Key findings from the GAO report include:
Current AI Uses: The federal government employs AI in 228 distinct ways, with nearly half of those uses initiated in the past year. Most applications (almost 70%) are science-related or aimed at enhancing internal agency management.
Secret AI Uses: While agencies disclosed about 70% of 1,241 active and planned AI use cases, over 350 applications were considered sensitive and not disclosed publicly. Some agencies, like the State Department, shared limited information about their AI applications.
Sensitive AI Applications: The Departments of Justice and Homeland Security reported a small fraction of use cases compared to NASA or the Department of Commerce, but these cases were identified as highly sensitive. The GAO previously warned about AI-powered facial recognition searches within law enforcement agencies without appropriate training requirements.
Government-Wide Standard: The report criticised the absence of a government-wide standard on AI purchases, emphasising the need for consistency. The lack of guidance from the Office of Management and Budget (OMB) has contributed to inconsistent policies across agencies.
OMB Delay: The OMB, responsible for aligning federal agencies' approach to AI procurement, missed the 2021 deadline to issue draft guidelines. The draft memo was eventually released in November 2023, contributing to inconsistencies in AI management practices.
Biden's Executive Order: President Joe Biden's recent executive order on AI safety requires developers of robust AI systems to share test results with the government. While OMB supports the GAO's recommendation for AI guidance, the delay has hindered the implementation of fundamental practices in managing AI.
The GAO report highlights the urgency for the U.S. government to establish coherent and responsible AI policies, addressing potential risks and ensuring alignment with crucial practices for the benefit of national security and public welfare.