By Steve Herne, CEO of Unlearn

The U.S. Food and Drug Administration (FDA) recently proposed a framework to evaluate the credibility of artificial intelligence (AI) used in drug development—a move that could reshape how clinical trials are designed and run. Here’s the bottom line: AI can accelerate the development of new treatments, but it is essential to assess and control the risk involved in such applications.

The FDA’s framework ensures that AI tools are evaluated based on their purpose, potential risks, and ability to deliver consistent results. For example:

• What problem is the AI trying to solve?
• How accurate and transparent are its predictions?
• How do we ensure the model’s results remain consistent over time.

By tailoring oversight to the specific risks posed by an AI tool, the FDA aims to build confidence among regulators, sponsors, and patients while maintaining the rigor required in clinical research.‍

What Sponsors and Researchers Should Know

Engage with the FDA Early

The new guidance encourages drug developers to start conversations with the agency before locking in a study design. That early feedback ensures that sponsors can align their trials with regulatory expectations and avoid costly delays.

‍

Be Transparent

Sponsors should be prepared to show how an AI model was created, how it’s tested, and how any ongoing updates are handled. By documenting these steps, you build credibility and trust with regulators and the broader scientific community.

‍

Evaluate Risk Thoughtfully

Consider where and how the AI model is used. For example, is the model making high-stakes decisions, such as determining patient safety protocols? Or is it supporting lower-risk tasks, like suggesting avenues for clinical research? Oversight will vary based on how directly the AI could impact trial outcomes and patient safety.

‍

The FDA’s guidance signals a pivotal moment for AI in clinical research. By setting clear expectations for transparency and robustness, the framework ensures that AI can enhance trial efficiency and outcomes without compromising safety or reliability.

‍

At Unlearn, we’re excited because this approach echoes what we’ve been advocating all along: AI, done right, can accelerate clinical development without introducing undue risk. In alignment with the FDA’s focus on risk-based evaluation, we’ve developed a clinical trial method—qualified by the European Medicines Agency in 2022—that rigorously manages risk and provides validation for AI models in their specific context of use. Our digital twins, which are AI-generated forecasts of control outcomes for clinical trial participants, are the foundational technology behind this method. For instance, we ensure methodological transparency by thoroughly documenting the internal workings of our AI models, including their architecture, training processes, and validation methods. These practices, combined with extensive pre-trial testing and early engagement with regulators, reflect our commitment to meeting and exceeding regulatory standards.

‍

If you’re curious about how digital twins can streamline clinical trials—whether for rare diseases, CNS indications like Alzheimer’s, or beyond—feel free to reach out. We’re here to share what we’ve learned, collaborate with industry leaders and regulators, and push the boundaries of what AI can do, responsibly.