DHS touts AI work and pilots a year after executive order
The Department of Homeland Security on Wednesday announced the successful tests of three generative artificial intelligence pilots, providing a progress report that comes as the White House declared many of the goals established in last year’s AI executive order complete.
The agency reported that U.S. Citizenship and Immigration Services received positive reviews from officers using generative AI-assisted training for testing, with pledges to continue to assess the technology in officer training. Similarly, Homeland Security Investigations, which evaluates global threats, said it will “continue to test and optimize the use of open-source models in supporting law enforcement investigations.”
The Federal Emergency Management Agency, meanwhile, appears to be in a preliminary phase with the technology. FEMA’s pilot involved the use of a large language model to help draft hazard mitigation plans, but the DHS component found 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.”
DHS continues to look at artificial intelligence overall, it said in Wednesday’s fact sheet. The agency has hired 31 new members of its AI Corps since February, and has also addressed national security concerns related to the technology through both a new board and a recent report into chemical and biological risks.
In a fact sheet also released Wednesday, the White House highlighted DHS’s work — along with the Defense Department — to use artificial intelligence to address vulnerabilities “for national security purposes and for civilian governmental organizations.”
In a compliance plan called for in an Office of Management and Budget memo on artificial intelligence, DHS also articulated some of its work on the technology. In that document, the agency said it had adopted a “default to open” approach to AI code and software — and committed to announcing which AI use cases it deemed rights- or safety-impacting, a special category with heightened scrutiny under new AI inventory rules.
In August, Customs and Border Protection published a list of documents related to its use of artificial intelligence, primarily several privacy threshold analyses.
