AI in Healthcare Is Advancing Faster Than Its Accountability Systems

On January 29–30, 2026, the GliaNet Alliance and the Coleman Institute for Cognitive Disabilities at the University of Colorado convened a two-day roundtable at CU Anschutz to examine how AI systems used in healthcare should be governed to protect patients and vulnerable populations. The gathering brought together clinicians, technologists, researchers, and governance experts to explore what duties intelligent systems should owe to the individuals whose care, communication, and daily decision-making they increasingly shape.

AI-enabled tools are already shaping clinical environments. Patient-facing chatbots assist with triage and communication. Cognitive-screening algorithms are used to detect early neurocognitive disorders. AI systems are increasingly embedded into electronic medical records and care workflows. Beyond software, sensor-driven tools and robotics are entering care settings. These systems influence decisions and patient experiences, yet they do not operate under a unified and enforceable duty to the individuals impacted by them.

Participants described what might be called a distrust gap. This does not reflect opposition to technological innovation. Rather, it reflects concern that governance, regulatory structures, and shared norms are not keeping pace with deployment. The comparison to the early development of social media surfaced repeatedly. In that era, rapid scaling and capital investment outpaced careful validation and institutional oversight. The consequences of that imbalance became clear only years later. Healthcare now faces a similar inflection point, though it remains early enough to shape outcomes more deliberately.

Existing regulatory frameworks address portions of the landscape but do not provide comprehensive accountability. HIPAA governs health data privacy, and the Food and Drug Administration oversees certain medical devices and software that meet specific criteria. Professional codes of conduct guide clinicians. However, many consumer-facing health technologies fall outside traditional regulatory categories. Large language models used for triage support or communication assistance may not qualify as regulated medical devices. Electronic Medical Records (EMR)-integrated decision tools can influence clinical pathways without carrying a clearly defined duty to patients. The result is a fragmented system in which responsibility is distributed across institutions, while patients experience these systems as integrated and unified.

The Coleman Institute's focus on cognitive disabilities brought particular clarity to the discussion. Older adults and individuals with cognitive challenges may benefit substantially from AI-enabled support tools. At the same time, these populations face heightened vulnerability to manipulation, fraud, and opaque algorithmic decision-making. Dominant AI business models often prioritize engagement and data extraction. Those incentives are poorly aligned with care environments in which vulnerability is central rather than peripheral. This misalignment raises a structural question about who is accountable for safeguarding patient interests when digital intermediaries mediate care.

The discussion moved beyond compliance-based approaches and toward structural governance design. Participants examined the possibility of establishing bifurcated duties for consequential AI systems. Under such a framework, AI systems used in health, finance, or other high-stakes domains would carry a legally enforceable duty of care. Organizations choosing to operate under a Net Fiduciary™ model would voluntarily adopt a duty of loyalty to the individuals they serve. This approach draws from established fiduciary regimes in financial and medical practice and reframes governance as part of system architecture rather than an external constraint.

In a healthcare context, this model could support the development of Personal Augmentation Agents that are aligned to individual patients rather than platforms. Such agents could assist individuals in interpreting test results, evaluating recommendations, and navigating AI-mediated decisions across institutional boundaries. This contrasts with large-scale agent systems optimized primarily for platform growth or market share. The distinction between platform-aligned systems and patient-aligned systems represents a meaningful governance choice.

The conversation also acknowledged the global dimension of AI governance. Regulatory approaches in the European Union classify certain systems as high risk, while other jurisdictions are developing distinct models. Health technology markets are international, and standards adopted in one region influence procurement and deployment in others. Fragmented governance without shared language increases uncertainty for developers and institutions alike.

Despite the urgency, participants emphasized that healthcare remains at an earlier stage of AI integration compared to other sectors. Institutional AI policies are still evolving. Procurement standards are being drafted. Research funding priorities are being shaped. Decisions made in the near term will influence how accountability, safety, and patient agency are embedded into systems that may become deeply integrated into care delivery.

The roundtable concluded with agreement on several next steps. The group identified research and policy gaps that require interdisciplinary collaboration, including validation frameworks for health-related AI systems and clearer intersections between regulatory regimes. Participants also emphasized the importance of leveraging procurement as a policy lever, since business contracts and institutional purchasing decisions shape incentives and system design in practice.

For the organizers, this convening reinforces the need to translate fiduciary principles into operational governance architecture. Healthcare presents one of the most consequential environments in which to test whether duties of care and loyalty can be embedded into intelligent systems as enforceable design choices. The Alliance will continue working with clinical leaders, researchers, and policymakers to advance frameworks that strengthen accountability and patient agency.

Organizations building, funding, regulating, or deploying AI systems in healthcare are invited to engage in this next phase of work. The governance choices made during this period will shape not only technological performance but also public trust in digital health systems for years to come.

About the GliaNet Alliance

GLIA Foundation, doing business as GliaNet Alliance, is a small nonprofit coalition that builds embedded fiduciary governance into commercial, digital public infrastructure (DPI), and governmental AI agent systems — an accountability layer designed to keep those systems loyally serving the best interests of the individuals and communities they touch. We work in four ways. Through the GliaNet Fiduciary™ Pledge — the program at the center of a coalition some two dozen organizations strong — companies commit to making client well-being the cornerstone of business decisions, protecting human agency as a fundamental right, and resolving conflicts of interest in the client's favor. Through research, we translate fiduciary principles into governance models institutions can implement inside systems already in operation. Through consultation, we work alongside those institutions to embed the governance layers into daily practice. And through policy engagement, we give lawmakers grounded input as the rules for AI agents are written. We work with partners seeking fully operationalized fiduciary governance for existing or planned pilots across the healthcare, education, workplace, civic, marketplace, and smart-city domains. Founded and led by former longtime Google director Richard Whitt, the GliaNet Alliance counts Consumer Reports among its members and has received initial philanthropic backing from the Omidyar Network.

About the Coleman Institute for Cognitive Disabilities

The mission of the Coleman Institute for Cognitive Disabilities is to promote the quality of life of people with cognitive disabilities and their caregivers by catalyzing advances in technology. The Institute supports research and development of technology on the four University of Colorado (CU) campuses in Boulder, Colorado Springs, Denver, and Anschutz for individuals living with cognitive disabilities and their caregivers. We are committed to providing scientific and technological leadership to strengthen the voice of our community. The Coleman Institute for Cognitive Disabilities was established in 2001 by the Regents of the University of Colorado. A private endowment and sustained annual contributions by our founding donors, Bill and Claudia Coleman, support the Institute's activities through the Coleman Colorado Foundation (CCF). CCF is a 501 (c) (3) public charity and is classified as a 509 (a) (3) supporting organization to CU that supports the Coleman Institute.