Traditional laboratory software operates on predictable programmed conditions, but AI-driven tools introduce complexities that existing regulations struggle to contain. Unlike conventional systems where errors are systemic and easily traced, AI models may produce unpredictable, patient-specific inaccuracies. These models can generate unsupported information or shift behaviors unexpectedly following minor updates, creating safety gaps that the 1992-era CLIA standards were never designed to manage.

ADLM President Dr. Stanley F. Lo argues that innovation must not outpace patient safety. The association recommends a risk-based approach that keeps laboratory directors at the helm of validation while ensuring AI is treated as an integral part of the total testing process. By avoiding a separate, redundant regulatory structure, the group aims to leverage the expertise already present in clinical labs to monitor performance, validate results, and maintain high-quality diagnostic standards in an era of rapid technological adoption.