Recently, the Department of Health and Human Services (HHS) Administration for Children and Families (ACF) issued new guidance on the use of predictive analytics in child welfare decision-making. The use of predictive analytics in these important decisions can have a significant impact on Tribal children and families. The National Congress of American Indians (NCAI) and the National Indian Child Welfare Association (NICWA) are sharing an important update about this new federal guidance and how Tribal Nations should engage with local, state, and federal governments to ensure that Tribal voices are part of the discussion about how predictive analytics are used when making decisions that could lead to the breakup of Tribal families.
Although this HHS guidance does not require any agency to adopt predictive analytics, it clarifies how jurisdictions can fund and integrate such systems into their technology modernization plans. On August 13, 2026, ACF announced that it was awarding $6 million in grants to 10 Tribal and state jurisdictions to support predictive analytics pilots for child welfare systems. That makes it even more critical for Tribal leaders to understand the potential risks for Tribal children, families, and Tribal sovereignty.
Why this matters
Predictive analytics in child welfare settings are promoted as tools to modernize decision-making and use data more effectively. At the same time, federal research has warned that predictive risk models can encode racial and ethnic bias and worsen existing disparities, including for American Indian and Alaska Native children, if they are not designed and governed with strong safeguards.
For background, see:
Avoiding Racial Bias in Child Welfare Agencies’ Use of Predictive Risk Models (HHS/ASPE).
Predictive Analytics in Child Welfare: An Introduction for Administrators and Policy Makers (ASPE).
Although predictive analytics may offer administrative benefits in some contexts, these same federal researchers have cautioned that, without careful safeguards, such systems may amplify existing disparities rather than reduce them. These risks are active and documented. For example, Oregon completely halted its screening algorithm after reviews showed it disproportionately flagged Black families. The pioneering Allegheny Family Screening Tool in Pennsylvania has faced U.S. Department of Justice (DOJ) Civil Rights Division scrutiny over concerns that it automates past inequalities. Meanwhile, states like Illinois have dropped predictive tools outright after finding them unreliable and ineffective.
Tribal data sovereignty and historical context
For generations, government systems and “expert” judgment were used to justify removing Indian children from their families and communities. Federal Indian boarding schools and other forced assimilation policies were advanced under the phrase “Kill the Indian in him, and save the man,” attributed to Richard Henry Pratt in 1892, which captured the goal of erasing Native identity in order to remake the child. That history and its continuing impacts are documented in efforts such as:
Following federal boarding school policies, state child welfare agencies took a larger role in intervening in Tribal child and family affairs. Prior to the passage of the Indian Child Welfare Act (ICWA) in 1978 [25 U.S.C. 1901 et seq.] an estimated 25-35% of Tribal children were removed from their homes by state and private agencies. Hearings before the House and Senate prior to ICWA’s passage also revealed that 85% of out-of- home placements of Tribal children were with non-Native families. NCAI has also affirmed the importance of protecting Indian children, families, and Tribal Nations by preserving the ICWA in Resolution MN‑23‑039, which recognizes ICWA as the minimum standard to prevent the unnecessary removal of Tribal children from their homes and communities.
Today, data about Tribal children and families are increasingly collected, stored, and analyzed in digital systems. Tribal data sovereignty means that Tribal Nations have the right to govern the collection, use, sharing, and interpretation of data about their citizens and communities, including in child welfare contexts. Any move to integrate predictive analytics with child welfare systems raises fundamental questions about who controls those data, how they are used, and how Tribal governments can protect their citizens.
Why Tribal Nations should pay attention
Predictive analytics systems are typically trained on historical administrative data. Those data reflect patterns of oversurveillance, overinvestigation, and overremoval that Tribal families have experienced for decades. When risk scores are generated from historically biased data, these tools can reinforce patterns of disproportionate surveillance and family separation that Tribal communities have experienced for generations. While the technologies are different from boarding schools, they risk perpetuating inequities rooted in the same history of government intervention in Tribal families. Even after ICWA’s passage, Tribal children continue to be overrepresented in state foster care systems, and Tribal families receive more intrusive interventions compared to other similarly situated families.
NCAI’s Resolution PDX‑11‑112 documented that, even decades after ICWA’s passage, federal data show persistently disproportionate foster care placements for Tribal children enrolled or eligible for enrollment and widespread state noncompliance with ICWA. Those same patterns of bias and noncompliance are likely to be reflected in the data used to train predictive analytics tools, unless strong Tribal‑centered safeguards are in place.
Key concerns include:
Systems may flag Tribal families as “high risk” more often if historical data show higher rates of prior involvement, without accounting for the structural causes of that involvement.
Systems that do not adequately address federal and state legal requirements or cultural values and norms that define the best interests of Tribal children and families may instead rely on one-size-fits-all definitions of those best interests.
Proprietary or opaque models for which commercial vendors claim confidentiality can make it difficult for Tribal Nations to understand how decisions are being influenced or to challenge unfair outcomes.
The threat of cross-system interoperability as local and state agencies increasingly combine child welfare data with homelessness, criminal justice, and public assistance databases. For Tribal families, this creates a massive, compounded web of surveillance that feeds directly into these algorithms, especially if Tribal consent is not central to design and oversight.
What NCAI and NICWA recommend
We urge our members to:
Center Tribal data sovereignty. Raise the need to recognize Tribal data sovereignty throughout participation in child welfare decision-making, and ICWA compliance, and discussions about predictive analytics in local, state, or federal child welfare systems.
Proactively engage local and state child welfare agencies. Decisions made by local and state agencies to adopt artificial intelligence (AI) methodologies, such as predictive analytics or risk modeling, directly impact Tribal children and families. This is true everywhere, including in urban areas and other regions without Tribal reservations or Tribal Nations within state boundaries. AI tools can be used for tasks ranging from helping case managers organize their case notes to influencing key decisions in child welfare.
Demand consultation. Ensure Tribal leaders and Tribal experts are consulted early and often before any tool is adopted.
Demand transparency. Transparency about how tools work, what data they use, and how they affect decisions across the child welfare process empowers Tribal Nations and supports sovereignty. This includes child protection investigations, the removal of children from their homes, placement decisions if a child is removed, access to family reunification and preservation services, and permanent placement outcomes such as reunification with birth parents, placement with relatives, or other permanent options.
Require equity safeguards. Demand third-party fairness audits to assess whether the algorithm discriminates against Tribal families. Additionally, require federal, state, and local agencies to disaggregate their performance data by Tribal affiliation, ensuring that disparities are not obscured by overall averages.
Mandate ICWA compliance by design. Demand that any predictive tools or algorithms used by federal, state, or local agencies be designed, monitored, and audited to ensure full compliance with ICWA requirements.
Assert the right to say no. Assert the right to decline participation in tools that increase surveillance or removal risks, and to negotiate data-sharing agreements that protect Tribal interests and values.
Resources
For members seeking additional background, the following resources offer more detail on the promises and risks of predictive analytics in child welfare:
