WASHINGTON — A processing error caused by an artificial intelligence tool used by Immigration and Customs Enforcement resulted in untrained or undertrained recruits being placed into field offices, sources told NBC News.
The AI tool was intended to identify applicants with prior law enforcement experience for ICE’s “LEO program,” a track requiring only four weeks of online training, the law enforcement sources told NBC. Those without such experience are required to complete an eight-week in-person course at the Federal Law Enforcement Training Center (FLETC) in Georgia.
| READ NEXT: Artificial intelligence and police leadership in 2026: From skepticism to stewardship
However, the tool allegedly mistakenly flagged applicants as qualified based on resume keywords such as “officer,” misidentifying titles like “compliance officer” or those expressing an aspiration to join ICE as equivalent to prior law enforcement experience.
As a result, officials said many applicants were routed into the shorter online training program without meeting the necessary background requirements. It remains unclear how many were improperly trained or deployed before the mistake was caught.
| DOWNLOAD: 26 on 2026: A police leadership playbook
The issue was identified in mid-fall, more than a month into ICE’s accelerated hiring push. ICE has since implemented manual resume reviews and recalled affected recruits to FLETC for proper training, according to the officials, who spoke anonymously due to not being authorized to discuss the matter publicly.
The hiring surge, backed by funding from the “One Big Beautiful Bill,” included $50,000 signing bonuses and was intended to meet a congressional mandate to add 10,000 new ICE officers by the end of 2025.
More than 2,000 ICE officers have been deployed to Minneapolis since late November, resulting in over 2,400 arrests, according to the report. The state of Minnesota has since filed legal action to remove the Department of Homeland Security from the region. DHS has not commented on the AI-related training issue.
