US All-in on AI, but Flying Blind Without a Comprehensive Plan, Says GAO

A recent federal oversight report indicates that the US government vastly lags behind on its goal to grow more reliant on AI. The fact that a government-wide standard hasn't been established for AI purchases can pose a major security risk, warned the Government Accountability Office (GAO).

The GAO, which happens to be the US' top watchdog for accountability, published a 96-page report on Tuesday that sheds light on the government's unpreparedness for the adoption of AI. The long-awaited survey reviews 23 agencies, seeking answers on their current and future plans for using AI.

Growing Adoption of AI in Government Agencies

The report in question is by far the most comprehensive effort by the US government to record the 200-odd ways in which AI and machine learning have found use in non-military agencies. It also covers the more than 500 ways in which the said agencies plan to use artificial intelligence in the future.

The GAO reports that AI currently has 228 distinct uses for the federal government. About 50% of these began during the past year, indicating a rapid adoption of AI across the government.

The report also suggests that about 70% of the current and ongoing uses of artificial intelligence in government agencies are either related to science or meant to enhance internal agency management.

Though the department listed as many as 71 different AI use cases, it told the GAO that only 10 of them could be publicly disclosed.

For example, the Department of Commerce told the GAO that it currently uses AI to track wildfires. Artificial intelligence has also been used to automate the process of keeping count of seals, walruses, or seabirds, the department reported.

At NASA, AI is used to track volcanic activity around the globe. The Department of Homeland Security confirmed that it uses artificial intelligence to identify border activities of interest" by evaluating camera and radar data using machine learning technologies.

The GAO report also reveals that AI is used by federal agencies in hundreds of secret ways. According to the survey, the agencies disclosed 70% of the 1,241 total active and planned use cases of AI publicly.

However, they refused to identify any more than 350 technology uses as they were deemed sensitive. Some agencies, such as the Department of State, were particularly secretive on this matter.

GAO Expresses Concern over a Lack of AI Standards and Policies

While the agencies are intent on using AI for a myriad range of applications, the GAO report warns that the government is lagging behind in creating policies for the responsible acquirement and use of AI technology from the private sector.

Until OMB issues the required guidance, federal agencies will likely develop inconsistent policies on their use of AI.GAO report

According to the reports, these policies might not be in line with key practices and potentially hinder American welfare and security.

The OMB was supposed to issue draft guidelines to the agencies by September 2021, under a 2020 federal law regarding the use of AI in the government. However, it only managed to publish the draft memo in November 2023, missing the deadline by over 2 years.

The GAO also warned in September that thousands of facial recognition searches had been conducted, with the help of AI, by law enforcement officials without appropriate training between 2019 and 2022. Experts have repeatedly cautioned about how excessive reliance on AI might lead to incidents of wrongful arrests due to mistaken identity.

The post US All-in on AI, but Flying Blind Without a Comprehensive Plan, Says GAO appeared first on The Tech Report.