Learn about the FDA, Health Canada, and MHRA GMLP guiding principles and how InferTrust™ creates verifiable records for training, validation, and monitoring act
In 2021, FDA, Health Canada, and the UK Medicines and Healthcare products Regulatory Agency (MHRA) jointly published ten guiding principles for Good Machine Learning Practice (GMLP). These principles establish expectations for how AI/ML-based medical devices should be developed, validated, deployed, and monitored throughout their lifecycle. InferTrust™ (Patent Pending) FDA Enforcement creates verifiable records that demonstrate adherence to these principles.
GMLP principles cover the full AI device lifecycle. Multi-disciplinary expertise: InferTrust™ decision records document the boundary between AI-generated outputs and clinical expert review, supporting the principle that AI device development requires diverse expertise. Good software engineering and security practices: InferTrust™ provides infrastructure-level security with device-bound cryptographic keys and tamper-evident record chains, going beyond application-layer security measures. Representative data and robust training: InferTrust™ seals training data fingerprints and validation results, creating verifiable records of data provenance and model training methodology. Clinically relevant performance on real-world data: Post-deployment decision records capture ongoing performance metrics across real patient populations, providing the real-world evidence that GMLP requires.
GMLP principles emphasize that deployed AI devices must be monitored for performance degradation, distributional shift, and adverse outcomes. InferTrust™ supports this by capturing every inference with sealed performance metrics, creating a longitudinal dataset that enables automated performance monitoring and early detection of model degradation. When performance thresholds are breached, the sealed record documents both the detection event and the subsequent corrective action.
While GMLP principles are not legally binding regulations, they represent the direction FDA expects AI device manufacturers to move. Organizations that implement infrastructure supporting GMLP adherence now will be well positioned when these principles evolve into enforceable requirements. InferTrust™ provides the verifiable documentation layer that transforms GMLP principles from aspirational guidelines into demonstrable practice.