DHS highlights hiring and project milestones on anniversary of Biden’s AI order

The Department of Homeland Security has “successfully tested” three generative artificial intelligence pilots and hired more than half of the experts needed to fill out its new AI Corps, according to an update published by the agency on Wednesday. 

The announcement came on the one year anniversary of President Joe Biden’s executive order on the safe and secure use of AI technologies, which established governmentwide guardrails around the adoption of the emerging capabilities. The White House said in a recent press release that all federal agencies have successfully completed their scheduled AI-related requirements over the past 12 months. 

DHS noted in a fact sheet that it was the first agency to issue an AI roadmap following Biden’s executive order, which it rolled out in March to outline its approach to using the new technologies. 

The guidance outlined three AI pilot initiatives DHS hoped to pursue in 2024, which were implemented within the Federal Emergency Management Agency, Homeland Security Investigations and United States Citizenship and Immigration Services.

The department said it tested all three programs by October and “gained valuable insights into the real-life impact of GenAI tools as well as their limitations.”

FEMA’s pilot project focused on helping state and local governments increase their resilience to emergencies by using a large language model to “generate draft plans customized to meet their needs and understand risks and mitigation strategies.”

DHS said the initiative helped the agency learn that “increasing user understanding of AI and receiving feedback directly from community users is an important first step to integrating GenAI into any existing process.”

Homeland Security Investigations — a law enforcement component within DHS that investigates domestic and global threats — also piloted the use of LLMs to summarize investigative reports and identify key words and phrases across different documents. DHS said the agency “found that open-source models provided the flexibility necessary to experiment and measure effectiveness.”

USCIS’s initiative leveraged GenAI capabilities to train immigration officers on how to interview refugee and asylum seekers and received the most praise from DHS. The department said the pilot received positive feedback from participating officers, particularly regarding its usability and accessibility. 

“Based on the success of the pilot, USCIS and DHS are looking at how GenAI can be used in other training scenarios as a supplemental tool to better prepare the next generation of DHS officers,” the department said.

Michael Boyce, director of the DHS AI Corps, previously noted during a conference in July that the USCIS pilot leaned into GenAI’s inconsistencies to better replicate actual interviews with asylum seekers.

“I also want them to hallucinate and I want them to be a little inaccurate because you're often, in real life, working with an interpreter and there's a lot of confusion and a lot of sort of dropped things, or things that don’t quite line up or make perfect sense,” he said. 

DHS also said it has hired 31 technology experts for its AI Corps since February, with the department planning to recruit a total of 50 specialists before the end of the year. The AI Corps hopes to deploy trained technologists across DHS to help with various initiatives and is modeled on the White House’s U.S. Digital Service.

The department said AI Corps experts have already “provided critical technical support and conducted extensive evaluations across multiple priority projects,” including helping Homeland Security Investigations create “a first-of-its-kind LLM-powered tool” for its pilot initiative.