An artificial intelligence error in processing applications sent many new recruits to US Immigration and Customs Enforcement into field offices without proper training as the agency raced to add 10,000 new officers to its force, according to two law enforcement officials familiar with the error.The AI tool used by ICE was tasked with identifying applicants with law enforcement experience for placement into the agency’s LEO programme, short for law enforcement officer, meant for new recruits who were already law enforcement officers and required four weeks of online training. Applicants without law enforcement backgrounds were required to take an eight-week in-person course at ICE’s academy at the Federal Law Enforcement Training Center in Georgia, which included courses in immigration law and handling a gun, as well as physical fitness tests, NBC news reported.“They were using AI to scan résumés and found out a bunch of the people who were LEOs weren’t LEOs,” one of the officials said.The officials said the AI tool sent people with the word “officer” on their résumés to the shorter four-week online training, including a “compliance officer” or people who said they aspired to be ICE officers. The majority of the new applicants were flagged as law enforcement officers, the officials said, but many had no experience in any local police or federal law enforcement force.Both officials said ICE’s field offices provided more training beyond what was provided at the academy or in the online course before officers were sent out onto the street, and that the officers singled out by the AI tool most likely received that training. The officials were not authorised to speak publicly and spoke to NBC News on condition of anonymity.The Department of Homeland Security did not respond to a request for comment. The officials said the AI mistake was identified in mid-fall, more than a month into the recruitment surge, and ICE immediately began taking steps to remedy the situation, including manual reviews of résumés of new hires.“They now have to bring them back to FLETC,” said one of the officials, referring to the Federal Law Enforcement Training Center.The officials said the AI tool was initially the mechanism used to categorise résumés. They were not sure how many officers were improperly trained, and it was not clear how many may have been sent out to begin immigration arrests.As the immigration agency surged agents into American cities, its enforcement tactics were increasingly questioned by local law enforcement, community groups and lawmakers following the shooting death of Renee Nicole Good in Minneapolis by ICE officer Jonathan Ross. Ross had more than 10 years of experience with ICE and would not have been subject to the AI screening for new recruits.The error highlighted the challenge of training such a large number of new recruits as ICE continued to ramp up operations to boost deportation numbers amid pressure from the White House. ICE also placed some new recruits into a training programme before they completed the agency’s vetting process, NBC News reported.In Minneapolis alone, more than 2,000 ICE officers were sent to the area to boost arrests, and they apprehended over 2,400 people since Nov. 29, DHS spokesperson Tricia McLaughlin said. Minnesota sued to try to remove DHS.ICE had a mandate to hire 10,000 new officers by the end of 2025 and offered new recruits $50,000 signing bonuses using the money Congress allocated under the One Big Beautiful Bill. One of the officials said that although ICE met the goal on paper, bringing back people who were misidentified for more training meant it did not successfully add 10,000 ICE officers on the street in 2025.