ICE AI system mislabels recruits, sends undertrained agents into field
An AI system used by U.S. Immigration and Customs Enforcement to vet recruits malfunctioned by flagging applicants as qualified law enforcement officers based on keyword matches in résumés rather than actual experience. Undertrained recruits without law enforcement backgrounds were sent into the field before the error was caught and manual reviews were implemented.
What happened
U.S. Immigration and Customs Enforcement deployed an AI system to identify applicants with law enforcement experience for its LEO program, a four-week online training shortcut. The system malfunctioned by flagging anyone with 'officer' in their résumé—including compliance officers and aspiring agents—as qualified law enforcement veterans.
The AI incorrectly classified most new applicants as experienced law enforcement officers even though many had zero experience with police or federal agencies. The error was not discovered until mid-fall, over a month into the recruitment drive, when ICE implemented manual reviews and recalled mislabeled agents for proper eight-week training at the Federal Law Enforcement Training Center in Georgia.
Reported impact
Undertrained recruits with no law enforcement background were sent into the field before completing the required eight-week training course in firearms handling, immigration law, and fitness tests.
Index analysis
The AI system matched keywords in résumés without verifying actual law enforcement experience, creating a false positive classification for candidates with incidental use of the word 'officer' in their backgrounds.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.