KFF warns AI could widen disparities
- KFF published an April 30 brief warning that AI is already spreading through care delivery, insurance, and patient advice — and could deepen inequities. - The sharpest signal is who uses AI now: 32% of adults, with higher mental-health use among uninsured, Black, and Hispanic adults. - That matters because biased training data, weak oversight, and digital access gaps can hard-code old disparities into new clinical tools.
Artificial intelligence in health care is no longer a future story. It is already doing billing, prior authorization, risk prediction, patient messaging, and increasingly, health advice itself. That is the setup for KFF’s new warning, published April 30: AI could help close gaps in care, but it could just as easily lock old inequities into faster, cheaper, more scalable systems. (kff.org) ### Where is AI already showing up? Pretty much everywhere. KFF walks through AI use in diagnosis, treatment planning, drug development, medical imaging, health monitoring, reimbursement decisions, and administrative work like scheduling, billing, and coding. Hospitals are using AI and p(kff.org)gement, and prior authorization. (kff.org) ### Why is this a disparities story? Because AI learns from the health system we already have — and that system is not neutral. If the training data underrepresent certain racial or ethnic groups, or reflect unequal access to care, the model can mistake missing care for lower need. Basic(kff.org)ale, inside tools that look objective. (kff.org) ### What does that look like in practice? It can show up in obvious ways and subtle ones. A model might work worse for groups that were less represented in the data. A chatbot might give weaker answers to people asking questions in another language. An automated workflow might steer peop(kff.org)ovals, delays, and follow-up. (kff.org) ### Why does mental health keep coming up? Because people are already using AI there, especially when the regular system is hard to reach. KFF’s March 2026 poll found that 32% of adults have used AI chatbots for health information or advice, and 16% have used them for mental-health infor(kff.org)ause the people leaning on AI most may also be the people most exposed if the tools are inaccurate, impersonal, or poorly governed. (kff.org) ### Why are people turning to AI at all? Speed, privacy, and access. Many users say they want immediate answers, want to look something up before seeing a clinician, or feel more comfortable asking sensitive questions privately. Some are using AI because they do not have a provider, cannot get an appo(kff.org)the health system left open. (kff.org) ### Can AI reduce disparities instead? Yes — but only if the deployment is deliberate. KFF argues that better outcomes depend on representative data, testing tools across groups, transparency about how models are built and used, and governance that catches unequal effects early. In plain English, heal(kff.org)ead end. (kff.org) ### What is the real warning here? The danger is not a single dramatic failure. It is quiet automation. If biased tools get embedded in intake, triage, claims, scheduling, or self-service advice, disparities can widen without any one person making an obviously discriminatory choice. That is what makes this report land. AI can make health care more efficient — but efficiency without equity just means unequal treatment happens faster. (kff.org) ### Bottom line KFF’s brief is basically a governance warning disguised as a technology story. AI may help expand access and reduce friction, but only if health systems measure who benefits, who drops off, and where human backup is still essential. Otherwise, the same communities already navigating the most barriers could become the testing ground for lower-trust, lower-accountability care. (kff.org)
