The Department of Homeland Security (DHS) released an updated version of its Artificial Intelligence (AI) Use Case Inventory this summer. At first glance, the changes seemed routine—some AI software programs were marked "inactive," and a new one was added. But upon closer examination, the removal of the AI use cases does not appear to indicate a retreat from—but an expansion of—those AI capabilities.

Taken alongside reporting from The Guardian and Wired, the update points to broader trends in immigration enforcement: deploying similar AI functions within larger vendor-run platforms and expanding into continuous surveillance systems that pull in and analyze far more information than before.

Two months earlier, in May, Immigration and Customs Enforcement (ICE) reported using 23 active AI software programs for immigration enforcement. By July, four of those had become "inactive," and one—Email Analytics for Investigative Data—was moved back to "implementation and assessment" phase for reconfiguration under a new system. At face value, this looks like a pullback in AI programs. However, recent reports show there is more going on. Many similar features are part of larger AI software platforms that generate automated decisions that are harder to oversee.