Regulating the Future: How the FDA Is Defining the Role of AI in Clinical Development

The U.S. Food and Drug Administration (FDA) has recently unveiled a new approval pathway (“plausible mechanism”) for personalized therapies targeting rare genetic diseases that could bypass the traditional randomized trial model. By allowing approvals based on data from a small patient pool, the FDA is signaling more than regulatory flexibility—it is embracing a data-driven, patient-centered approach to clinical trials.

This news reflects a broader transformation in clinical development. Therapies are becoming increasingly tailored, and both trial design and the way evidence is generated must evolve, with artificial intelligence (AI) essential in enabling that progress. As AI becomes more deeply embedded, it offers opportunities to improve precision, streamline processes, and strengthen patient-centricity. 

Yet with such innovation comes a shared challenge for regulators and biopharmaceutical companies: ensuring AI is used safely, effectively, and transparently.

In 2025, several initiatives have demonstrated how the FDA is responding to this changing landscape. The agency released draft guidance for the responsible use of AI across the drug product life cycle, including clinical trials, giving sponsors a clearer framework for implementation. This was followed by the launch of Elsa, the FDA’s own agency-wide AI tool. Elsa promises to modernize and accelerate protocol reviews by using automation and advanced analytics to shorten evaluation timelines.

Together, these initiatives and the FDA’s exploration of new approval pathways show that regulators are not only setting guardrails for AI but also recognizing its transformative potential in clinical development.

The evolving role of AI in clinical development

Despite ongoing efforts to improve it, around 90 percent of clinical drug development still fails. Two persistent contributors to underperforming trials are protocol design issues and inadequate investigator site selections. To address these pain points, sponsors are increasingly turning to AI and clinical data analytics to identify high-enrolling investigator sites for each disease, eliminate non-enrolling sites and reduce unnecessary protocol amendments. This approach allows the industry to capitalize on growing volumes of historical and real-world data to bring greater precision to trials.

AI-driven applications and models are also improving the efficiency and transparency of clinical trials and regulatory submissions. The FDA is using AI and machine learning to process and analyze large datasets, support the development of trial endpoints, and evaluate outcomes. By organizing and interpreting complex data more effectively, these tools help produce more structured approval packages for reviewers, strengthening regulatory trust and overall trial integrity.

AI is further enabling innovations such as patient “digital twins,” which can serve as simulated control arms. Digital twins address ethical concerns around placebos, support trials where traditional control groups are impractical, and reduce protocol amendments by allowing sponsors to understand patient characteristics before a study begins. 

This particular evolution sits squarely within the broader regulatory shift referenced earlier. The FDA’s new plausible mechanism pathway allows approvals for rare diseases based on data from very few patients, reflecting a willingness to rethink evidence standards when large randomized trials are not feasible. 

Although digital twins introduce new regulatory considerations, and the FDA has not yet issued dedicated guidance, referencing them only within broader modeling and simulation materials, digital twins align closely with this move toward flexible and data-efficient evidence generation.

The regulatory landscape: where things stand today

The FDA’s draft guidance on using AI to support regulatory decision-making represents a significant step toward addressing the challenges of integrating emerging technologies across a drug product’s life cycle. 

Central to the guidance is a seven-step framework for assessing AI model credibility, ensuring systems are robust, reliable, and aligned with regulatory expectations. This structured, risk-based approach helps both regulators and developers determine whether an AI model can be trusted to support critical decisions.

Looking ahead, the FDA strongly encourages early engagement from sponsors to identify potential challenges, clarify expectations, and inform credibility assessments—particularly when incorporating new elements such as digital twins into a trial. 

Early discussions allow sponsors and regulators to align on questions of data quality and model validation. This proactive approach strengthens regulatory confidence in AI-enabled tools and helps sponsors design trials that align with regulatory priorities from the outset.

Building regulatory trust: what sponsors must do

Regardless of what specific legislation is ultimately enacted, AI will only gain regulatory acceptance if sponsors demonstrate scientific rigor and ethical responsibility in how AI-driven platforms and models are deployed. 

Three steps are essential:

1. Define and validate 

AI systems must clearly articulate their context of use, demonstrate validation on representative data, and remain interpretable and reproducible.

2. Ensure ethical data governance

Given reliance on sensitive patient data, sponsors must follow governance frameworks covering data sourcing, ownership, and ethical use.

3. Document model development

As AI systems take on greater roles in trial design and results interpretation, auditable documentation of model development, decision frameworks, and validation criteria is crucial.

Charting the next five years of AI-driven clinical trials

As AI-driven platforms and models continue to gain ground in clinical development, the next five years will be pivotal for establishing regulatory frameworks that balance innovation with patient safety and data integrity. 

Continued collaboration between the FDA and industry stakeholders, supported by ongoing feedback loops, will be essential to ensure emerging technologies deliver on their promise of faster and more effective trials. 

Sponsors who implement AI tools responsibly will be best positioned to design patient-centric trials, accelerate cycle times, and build lasting regulatory trust.