In President Trump’s executive order on foster youth, issued in November of 2025, he instructed the Department of Health and Human Services to take actions that would:
“Expand states’ use of technological solutions, including predictive analytics and tools powered by artificial intelligence, to increase caregiver recruitment and retention rates, improve caregiver and child matching, and deploy federal child-welfare funding to maximally effective purposes and recipients.”
Without an indication of any forthcoming financial support for it, the Administration for Children and Families (ACF) made clear in an issue brief that it views predictive and AI tools as a positive part of child welfare’s future.
“When paired with strong practice protocols, workforce training, and transparent governance, these tools provide a clearer, more comprehensive view of risk and case complexity — precisely when that information is most needed,” the brief says. “States that adopt and thoughtfully implement these tools will be better positioned to protect children, serve families, and deliver a more responsive and effective child welfare system.”
ACF, which held an in-person convening about predictive and AI technology last year, did note the belief that they bake in bias that already exists in data about citizens, especially low-income families, and engrain that bias in opaque tools. The greatest concern is reserved for the use of such instruments around decisions to investigate a report of maltreatment, or to remove a child into foster care.
But, the brief states, the same critique could be made of the “traditional, checklist-based risk assessment tools” used by many systems today. So the debate should not be about the potential for bias, but instead focus on “how openly those biases are examined and whether the tools in use — old or new — are held to clear standards for transparency, accuracy, and consistency.”