Grassroots Action Center
Insurers are investing heavily in AI review tools as a cost-saving measure, while shifting financial and administrative burdens to patients and healthcare professionals. This administrative burden is resulting in practice consolidation at an accelerated rate, with only about 42% of physicians remaining in independent, physician-owned practices today and the number of independent rural physicians falling 43% between 2019 and 2024. This leaves patients with fewer choices and longer drives and wait times for care. The DNAI Act does allow for insurance plans and issuers to use AI as a tool to enhance efficiency and legitimately reduce administrative costs, but it prohibits them denying care by substituting AI for the clinical judgment of a licensed professional.
The DNAI Act helps protect patients by:
- ensuring AI only plays a supportive role—prohibiting AI from issuing or making benefit determinations involving medical judgment and requiring that medical-necessity denials be made by a licensed health care professional with appropriate expertise.
- protecting clinical judgment—ensuring that it is not issued by an AI system and only by a licensed healthcare professional who is acting in the scope of their license.
- increasing transparency of AI utilization—requiring disclosure when AI is used in review, including AI-generated outputs in the administrative record, and making documentation available to patients upon request.
- reinforcing mental health parity—treating AI use in mental healthcare as a treatment limitation to be subject to evaluation under the Mental Health Parity Law, helping prevent disproportionate algorithmic scrutiny of mental health and substance use disorder care.
The DNAI Act could help protect Mental Health Counselors’ clinical judgment and reduce barriers to medically necessary mental healthcare. Overall, the DNAI Act would ensure that AI remains a tool to support healthcare—not a replacement for the expertise and clinical judgment of licensed mental health professionals.
