From Medical Devices to Drug Development: Advancing Responsible AI in Healthcare

Artificial intelligence (AI) is increasingly transforming healthcare, from clinical decision-making to drug discovery and development, heightening the need for clear regulatory guidance to ensure safe, reliable, and patient-centered use. Two major international frameworks – the FDA – led Good Machine Learning Practice for Medical Device Development: Guiding Principles (October 2021) and the EMA-driven Guiding Principles of Good AI Practice in Drug Development (January 2026) – illustrate how authorities are shaping responsible AI across healthcare. This blog compares these frameworks to clarify how evolving regulatory expectations are influencing the development and deployment of AI-enabled healthcare solutions

A Shared Commitment to Patient Safety and Quality

Both guidance documents are found on a common objective: ensuring that AI technologies support high standards of quality, safety, and clinical effectiveness.
The 2021 Good Machine Learning Practice (GMLP) framework emphasizes robust engineering practices, representative data, statistically sound and clinically relevant validation, and ongoing post-market performance monitoring for AI-enabled medical devices. Similarly, the 2026 drug development principles reinforce that AI must strengthen, rather than replace, established scientific and regulatory standards.
In both cases, regulators view AI as a powerful tool that must operate within existing quality and compliance frameworks.

Different Contexts, Complementary Approaches

While aligned in purpose, the two frameworks address distinct areas of healthcare innovation.
The GMLP principles focus on AI-enabled medical devices that interact directly with healthcare professionals and patients. As a result, they emphasize usability, clinical workflow integration, and real-world performance.

In contrast, the drug development principles apply to AI systems used across the entire pharmaceutical life cycle, including nonclinical, clinical, post-marketing, and manufacturing phases. This broader scope reflects the complexity of medicine development and the long-term impact of AI-driven decisions.

Strengthening Human-Centric and Ethical Design

A key evolution between the two frameworks is the increased focus on ethical and human-centered design.
The 2026 guidance explicitly highlights “human-centric by design,” emphasizing alignment with ethical values and patient interests. While the 2021 framework addresses human-AI interaction and interpretability, the newer principles extend this focus to broader societal and ethical considerations.
This shift reflects growing recognition that responsible AI must balance technical performance with transparency, fairness, and trust.

Advancing Risk-Based and System-Level Oversight

Effective risk management is central to both documents.
The GMLP framework emphasizes technical risk controls across the total product lifecycle, including dataset independence, robust cybersecurity practices, statistically sound performance testing, and post-deployment monitoring to mitigate bias and model drift, thereby supporting reliable clinical use.
The 2026 principles adopt a more holistic, risk-based approach that considers the entire AI-enabled system, including organizational processes, governance structures, and human oversight. Validation and monitoring are proportionate to the intended use and potential impact of each system.
This progression demonstrates a shift from model-focused evaluation to comprehensive system governance.

Enhancing Data Governance and Traceability

High-quality data is fundamental to trustworthy AI.
The 2021 guidance emphasizes representative datasets, validated reference standards, and careful separation of training and testing data. These practices promote generalizability and reduce bias.
Building on this foundation, the 2026 framework introduces more extensive requirements for data provenance, documentation, and traceability, aligned with Good Practice (GxP) standards. This supports auditability, regulatory confidence, and long-term reliability.

Lifecycle Management and Continuous Improvement

Both frameworks recognize that AI governance extends beyond initial deployment.
The GMLP principles require continuous monitoring and controls for retraining and performance degradation. This is particularly important for adaptive and learning systems.
The drug development guidance further integrates AI oversight into formal quality management systems, requiring scheduled reviews and periodic re-evaluation. This ensures that performance, safety, and compliance are maintained throughout technology’s lifecycle.

Promoting Transparency and Stakeholder Engagement

Clear and accessible communication is a shared priority.
Both documents emphasize providing users with essential information regarding intended use, limitations, performance characteristics, and updates. The 2026 framework places additional focus on plain-language communication and patient accessibility, reflecting the expanding role of patients and the public in healthcare decision-making.

Transparent communication supports trust and informed adoption of AI technologies.

Implications for Healthcare and Life Sciences Organizations

Together, these regulatory frameworks demonstrate how expectations for AI governance are evolving. Organizations developing AI-enabled solutions must now consider not only technical performance, but also ethical design, data governance, quality systems, and stakeholder communication.
Key success factors include:

• Integration of multidisciplinary expertise
• Robust risk-based governance
• Comprehensive data management
• Continuous lifecycle oversight
• Commitment to transparency and accountability

By aligning innovation with these principles, organizations can build AI technologies that deliver sustainable value and regulatory confidence.

Conclusion

The 2021 GMLP principles established a strong technical foundation for AI in medical devices. The 2026 drug development guidance builds upon this foundation, introducing enhanced governance, ethical considerations, and system-level oversight.
Together, they reflect a maturing regulatory landscape – one that supports innovation while prioritizing patient safety, scientific rigor, and public trust.
As AI continues to shape the future of healthcare, these frameworks provide a clear pathway for developing technologies that are not only advanced, but also responsible and reliable.