GAO Identifies Error in Homeland Security Department’s AI Use Case Inventory

The Government Accountability Office is questioning the reliability of the Department of Homeland Security’s artificial intelligence use case inventory after identifying an incorrect entry.

DHS officials confirmed to the watchdog that one cyber use case, a predictive model for sharing cyberthreat information, was wrongly added to the initial list, FedScoop reported.

GAO wrote in a new report that the department admits that it does not verify whether assessed use cases are AI. The watchdog explained that DHS cannot ensure that its inventory is accurate until it adds such checks to its review process.

In the report, GAO said the issue came up in part because DHS did not fully adopt a prescribed AI policy. The department implemented 11 practices in the 2021 AI Accountability Framework to support its use of AI for cybersecurity.

Officials from DHS’ Chief Technology Officer Directorate reasoned that their verification efforts relied on components and existing information technology governance efforts.

The department agreed with GAO’s eight recommendations to remedy the issue, including an expanded AI review process, new accuracy checks for inventory submissions and full implementation of the 2021 AI Accountability Framework.