How AI Transforms Regulatory Submission: Current Clinical Implementation and Future Prospects

ABSTRACT

Artificial Intelligence (AI) is transforming drug development and regulatory submission by enabling advanced data analytics, predictive modeling and intelligent decision support systems. Beyond efficiency gains, AI establishes a translational bridge between model‐informed drug development (MIDD) and clinical implementation, turning regulatory evidence into actionable insights that enhance therapeutic precision and patient outcomes. This perspective paper explores AI's current applications, regulatory integrations, and future prospects in accelerating data‐driven, patient‐centered drug development.

1. Evolution of AI in Healthcare and Drug Development

Advanced AI technologies, including Machine Learning (ML), Deep Learning (DL), Natural Language Processing (NLP), and Large Language Models (LLMs), are rapidly transforming healthcare and drug development (Figure 1). ML enables data‐driven learning without explicit programming, while DL improves predictive accuracy through multilayered neural networks. NLP and LLMs interpret unstructured biomedical text, summarize clinical data, and extract insights from diverse datasets, thereby supporting evidence generation across the development continuum.

FIGURE 1.

Schematic of AI landscape in 2025.

Recent innovations in transformers and diffusion architectures, state‐space and liquid foundation models, and multimodal and agentic retrieval‐augmented systems have significantly expanded AI's analytical, predictive and generative capabilities. These advances underpin the emergence of AI agents capable of autonomous reasoning and task execution, accelerating scientific discovery. While these innovations demonstrate remarkable potential, AI hallucinations, where AI systems generate convincing but factually inaccurate information, represent a significant challenge in current AI development. Hallucinations stem from the complex probabilistic nature of LLMs and can occur when AI systems confidently generate responses that sound plausible but lack factual grounding. AI hallucinations are outside the scope of this article.

AI has also revolutionized healthcare, particularly drug development. These technologies are fundamentally changing pharmaceutical research across the entire drug development pipeline. LLMs now rapidly analyze vast medical literature, predict patient outcomes, and enable truly personalized treatment approaches. This technological evolution facilitates more efficient, targeted clinical trials with improved success rates and accelerated timelines. A recent study by the Tufts Center for the Study of Drug Development, in collaboration with the Drug Information Association and 16 biopharmaceutical companies and contract research organizations, mapped the current use of AI in drug development [1]. The study highlighted key functional areas of AI implementation such as clinical trial design and planning, literature review, protocol design, and educational content development. Yet, challenges remain, including data quality, trust in AI‐generated outputs, legal and intellectual property concerns, ethical and privacy matters, and the need for further validation.

2. Regulatory Considerations for Using AI in Drug Development

This paper examines the current regulatory landscape for integrating AI technologies from the drug developer's perspective, focusing on how these tools can generate robust evidence for regulatory submissions. As AI use expands from optimizing clinical trial design to real‐world evidence (RWE) analysis, sponsors are exploring innovative approaches to strengthen regulatory submissions, quality, comprehensiveness, and speed while maintaining rigorous scientific standards of validation, transparency, and reliability.

AI is transforming modeling and simulation workflows across drug development, from addressing translational PK/PD and biomarker modeling questions that guide first‐in‐human studies, to quantifying population heterogeneity in disease progression and exposure–response, optimizing late‐phase trial design through subpopulation identification, and utilizing synthetic controls or digital twins. For high‐risk regulatory applications, AI models may require detailed submissions, including model architecture, training logs, and data processing pipelines as part of clinical trial documentation [2]. Protocol considerations must address AI's influence on trial design, decentralized elements, and decision‐support functions. AI models used for data transformation, analysis, or interpretation in clinical trials must adhere to established statistical guidelines. Early‐phase trials leveraging AI‐driven insights are generally lower risk, but high‐stakes applications such as treatment assignment or dosing demand rigorous validation and sensitivity analyses to mitigate biases.

To harness AI's full potential while maintaining rigorous safety, efficacy, and quality standards would require strategic cooperation among regulatory bodies, industry, and research institutions. Public–private partnerships (PPPs) such as the Critical Path Institute (C‐Path), the Foundation for the National Institutes of Health (FNIH), TransCelerate Biopharma, the Alliance for AI in Healthcare (AAIH), the Coalition for Health AI (CHAI), and others are helping bridge this gap. We call on these organizations to prioritize the training and adoption of these fast‐evolving technological advancements because they hold immense potential to transform the drug development process. Such PPPs are vital for translating regulatory expectations, rapidly developing implementation frameworks, validating AI methodologies, and enabling global harmonization. PPPs will facilitate pre‐competitive collaboration to define standards and best practices for AI validation, bias assessment, and transparency, aligning global regulatory requirements and streamlining AI adoption globally. By strengthening trust and transparency, these collaborations can help translate the rapid advancements in AI into actionable, regulatory‐ready evidence that accelerates drug development and safeguards public health.

3. Regulatory Decision Making in the Era of AI

The landscape of drug development is rapidly evolving with the integration of AI. These tools have become instrumental in harnessing vast, multimodal data generated from real‐world sources, allowing for deeper insights and more efficient decision‐making processes. AI enhances the efficiency of drug development tools (DDTs) by rapidly analyzing complex datasets, identifying promising candidates, and optimizing decision pathways. This capability is particularly valuable in the development of specialized tools such as clinical trial simulation and enrichment strategies, which improve trial design and patient selection. The synergistic integration of AI and DDTs marks a significant advancement in the pursuit of innovative and targeted therapeutic solutions. Regulatory bodies worldwide are adapting to this technological shift by updating guidelines and frameworks to accommodate the novel capabilities of AI [3]. This evolution is crucial for maintaining rigorous safety standards while fostering a regulatory environment that enables AI‐driven innovations, ultimately expediting the approval of effective and safe therapies. While AI excels at analyzing complex data, expert oversight remains essential to interpret outputs within clinical and regulatory contexts. Human‐in‐the‐loop approaches ensure that AI‐generated insights align with scientific standards, ethical considerations, and patient safety priorities.

In early 2024, the FDA established the Quantitative Medicine Center of Excellence to facilitate and coordinate the application of quantitative solutions, including AI [4]. This specialized center coordinates initiatives across both internal review divisions and external stakeholders, concentrating on three strategic priorities: regulatory science, enterprise planning and coordination, and education and ecosystem development. In another landmark development, the FDA announced the completion of its first AI‐assisted scientific review pilot in May 2025 and introduced an ambitious timeline for an agency‐wide AI rollout [5]. Similarly, the European Union's AI Act (Regulation 2024/1689) introduced a risk‐based framework addressing transparency, human oversight, and bias mitigation for high‐risk medical AI systems. While these requirements may extend development timelines, they are essential to ensure the safe, ethical, and effective use of AI.

4. AI Use Cases in Regulatory Submissions

The integration of AI into regulatory submissions is gaining momentum, offering innovative approaches to data analysis, patient selection, trial design, and safety evaluation. These use cases span multiple regulatory pathways and jurisdictions, illustrating the growing acceptance of AI‐based methodologies by global health authorities. Table 1 presents representative examples of AI applications that have been incorporated into regulatory submissions, highlighting their relevance and regulatory context. The use cases represent a continuum of AI applications, from modeling virtual control arms and quantifying disease activity, to identifying patient subgroups, analyzing real‐world drug use, and predicting toxicity risks. Collectively, they demonstrate how data‐driven modeling and RWE approaches can accelerate regulatory decision‐making and improve patient safety and treatment efficacy across the drug lifecycle.

TABLE 1.

AI use cases in regulatory submissions.

Use case	Description	Regulatory submission type	Contribution
Digital twin and prognostic covariate adjustment [6]	The EMA‐qualified methodology uses deep learning to generate digital twins that simulate placebo or control arms in clinical trials. By adjusting for baseline prognostic covariates, it enhances statistical precision and reduces the need for traditional control groups	EMA Qualification Opinion (EMADOC‐1700519818‐907,465)	Demonstrates how AI can simulate control arms, reducing reliance on traditional trial designs
Determining disease activity in NASH/MASH clinical trials [7]	An EMA‐qualified AI model analyzes digitized liver biopsy images to quantify histological features of NASH/MASH. This approach provides reproducible, quantitative measurements that improve the consistency of disease activity assessment across trials	EMA Qualification Opinion (EMADOC‐1700519818‐1,761,332 Corr.1)	Demonstrates the use of AI for automated, objective pathology assessment to improve endpoint consistency
Patient population selection [8]	A ML model identified COVID‐19 patients most likely to benefit from nakinra by analyzing clinical and laboratory data to detect hyperinflammatory signatures. This predictive approach guided patient selection under FDA Emergency Use Authorization, supporting targeted treatment strategies	FDA EUA 109	Illustrates how ML supports adaptive trial design and patient stratification for targeted therapies
Exploring drug utilization from RWD [9]	Symphony Health's Metys platform applied artificial intelligence and machine learning to integrate and analyze large‐scale real‐world healthcare data, including prescription, diagnostic, and treatment records for phenobarbital sodium. These AI‐driven analytics identified utilization patterns and treatment outcomes in neonatal and preterm infants, providing regulatory‐grade real‐world evidence that supported FDA evaluation of clinical relevance and safety	FDA NDA 215910	Demonstrates how AI‐powered RWE analytics enhance regulatory insight into real‐world treatment use, safety, and benefit–risk assessment
Predicting liver and renal toxicity [10]	ML models were employed to predict potential adverse effects of remdesivir and its metabolites. These predictions were part of a broader computational toxicology analysis to assess the safety profile of the drug	FDA NDA 214787	Demonstrates the application of AI for in silico toxicology and predictive safety analysis

The rapid evolution of AI and its integration into regulatory decision‐making presents challenges for sponsors, who must navigate regulatory uncertainty, maintain compliance, and validate AI‐driven methodologies. As regulatory frameworks continue to develop, sponsors must address these complexities while ensuring the safety and efficacy of AI‐powered drug development tools.

As highlighted earlier, PPPs offer an ideal framework for addressing AI integration challenges in healthcare regulation and drug development. C‐Path, a nonprofit partnership with the FDA, stands at the forefront of this evolving landscape. Its Regulatory Science Program consolidates insights across consortia, enabling productive dialogue with regulatory authorities, while its Quantitative Medicine Program implements AI solutions for complex data integration and analytics, enhancing clinical trials, biomarker discovery, and regulatory assessments. C‐Path's partnership model creates a neutral collaborative space where industry experts, academic researchers, and regulators can establish best practices, minimize risks associated with AI innovations, and develop practical regulatory applications for new medical products. This collaborative approach maintains compliance with emerging guidelines while building a sustainable ecosystem for AI‐driven pharmaceutical development that ultimately benefits patients.

5. Translational and Clinical Implications

Leveraging AI into regulatory frameworks has direct clinical relevance. By linking MIDD with patient‐level outcomes, AI can inform dose optimization, adaptive trial designs, and therapeutic stratification across diverse populations. The same analytical frameworks can later guide clinical dose adjustments, risk assessment, safety monitoring, and treatment guideline refinement, thereby supporting precision medicine initiatives. As regulatory agencies adopt AI‐driven solutions, the resulting feedback will increasingly shape evidence‐based standards that inform both clinical protocol design and real‐world implementation. In this way, AI will serve as a translational bridge between regulatory assessment and patient‐centered care, accelerating the feedback loop between research, regulation, and clinical practice.

6. Future Directions and Translational Outlook

The integration of AI into regulatory frameworks represents a pivotal shift in pharmaceutical development and healthcare delivery. As these technologies continue to evolve, collaborative approaches involving regulatory bodies, industry partners, academic institutions in PPPs will be essential to establish robust standards while fostering innovation. C‐Path's PPP model integrates human subject matter experts into AI workflows through expert review, transparent model evaluation, and co‐development of best practices, ensuring AI enhances rather than replaces critical human judgment and accountability in regulatory decision‐making.

Building on this foundation, future scientific efforts should increasingly focus on these areas in a way that can withstand regulatory scrutiny. By addressing scientific challenges related to data privacy, algorithmic transparency, model validation, and regulatory compliance, stakeholders can create a balanced ecosystem that accelerates drug development without compromising safety or efficacy. This coordinated effort will ultimately transform patient care by enabling more personalized treatments, efficient clinical trials, and data‐driven regulatory decisions that will expedite the delivery of effective and innovative therapies for patients.

Looking ahead, embedding AI‐driven regulatory‐grade solutions within the broader translational continuum will help seamlessly connect discovery, development, clinical implementation and RWE generation facilitating a fully integrated patient‐centered translational science ecosystem.

Funding

Critical Path Institute is supported by the Food and Drug Administration (FDA) of the Department of Health and Human Services (HHS) and is 56% funded by the FDA/HHS, totaling $23,740,424, and 44% funded by non‐government source(s), totaling $18,881,611. The contents are those of the author(s) and do not necessarily represent the official views of, nor an endorsement by, FDA/HHS or the U.S. Government.

Conflicts of Interest

The authors declare no conflicts of interest.

Podichetty J. T., Bauer A.‐M., Xu R., et al., “How AI Transforms Regulatory Submission: Current Clinical Implementation and Future Prospects,” Clinical and Translational Science 18, no. 12 (2025): e70434, 10.1111/cts.70434.