Navigating FDA’s Proposed Guidance on AI and Non-Animal Models: Safeguarding Innovation in Drug Development

In April 2025, the U.S. Food and Drug Administration (FDA) launched a landmark steerage titled “Roadmap to Reducing Animal Testing in Preclinical Safety Studies,” outlining its dedication to advancing New Approach Methodologies (NAMs) — together with in silico fashions, organoids, and different non-animal options. This steerage encourages sponsors of Investigational New Drug (IND) functions to undertake scientifically credible options to animal research, marking a shift towards extra human-relevant, Artificial Intelligence (AI)-integrated platforms for regulatory submissions.

Earlier, in January 2025, the FDA launched a companion piece of draft steerage titled “Considerations for the Use of Artificial Intelligence to Support Regulatory Decision-Making for Drug and Biological Products.” We mentioned this FDA steerage in element right here: AI Drug Development: FDA Releases Draft Guidance.

Briefly, this steerage focuses particularly on the usage of AI throughout the drug improvement lifecycle. It introduces a risk-based framework primarily based on the mannequin’s context of use and outlines the data that have to be disclosed about mannequin structure, knowledge governance, life cycle upkeep, and potential scientific impression — notably the place the mannequin influences affected person security or drug high quality.

A rising variety of real-world examples verify that AI-powered NAMs have gotten actuality and display profitable integration of AI and life science. Notably, the UVA/Padova Type 1 Diabetes Simulator, a physiologically detailed silico platform, has already been accepted by the FDA and used to help regulatory clearance of steady glucose monitoring (CGM) gadgets. This is a tangible sign that digital physiological techniques can now kind a part of the evidentiary basis for FDA approval.

FDA’s inside embrace of AI reinforces its regulatory messaging and alerts long-term institutional dedication to digital instruments. In explicit, the FDA publicized its new makes use of of AI instruments for inside operations, together with a generative AI software, “Elsa.” The FDA indicated that it could incorporate these instruments broadly all through the company by the top of June 2025.

Together, these FDA efforts mark a big evolution: AI and others in silico platforms are actually not solely permissible however more and more central to preclinical and scientific decision-making, together with regulatory assessment. With that recognition comes a twin crucial for sponsors and builders to display credibility to regulators, and safeguard innovation with strong IP technique. NAMs considerably improve corporations’ danger publicity with respect to each privateness and confidentiality issues.

The sections that comply with discover how corporations and their authorized groups can reply strategically — by leveraging rising FDA frameworks, defending improvements in AI modeling, and guaranteeing knowledge governance practices are aligned with each regulatory and IP targets.

What Are NAMs?

NAMs discuss with non-animal methods for evaluating drug security and efficacy. These embody in silico fashions, microphysiological techniques, organoids, computational toxicology platforms, and AI-driven scientific simulations. NAMs supply advantages in velocity, price, and moral soundness, whereas aligning extra carefully with human biology.

Despite speedy progress, FDA’s formal acceptance of AI-based NAMs stays restricted. The UVA/Padova Type 1 Diabetes Simulator stays the one broadly cited instance of an in silico mannequin that has been used efficiently in FDA regulatory submissions — particularly in evaluating CGM gadgets. Its acceptance underscores the potential for regulatory pathways for complicated physiological simulators.

However, the ecosystem of NAMs is quickly increasing. While not all have but achieved full FDA qualification, a number of promising AI- and data-driven platforms are into consideration:

• Organ-on-a-Chip and Organoid Models: Patient-derived organoids and organ-on-a-chip techniques are being validated to simulate tissue-specific responses. For instance, intestinal organoids have been proven to foretell off-target toxicity for T-cell therapies.
• AI-Based Computational Toxicology: AI fashions educated on large-scale toxicology databases are being developed to foretell opposed outcomes and are supported by FDA by collaborative validation initiatives.
• In Silico Clinical Trials (ISCTs): Computational affected person fashions are being piloted to simulate scientific trial outcomes, notably for system testing, with outlined workflows for mannequin validation and uncertainty evaluation.
• Physiologically Based Pharmacokinetic (PBPK) Models: These simulate drug distribution and metabolism and are into consideration as partial replacements for animal testing in pharmacokinetic profiling.
• Synthetic Control Arms: AI-derived digital affected person cohorts are gaining traction as replacements for placebo teams in trials, serving to scale back the variety of actual affected person individuals.
• Wearable-Integrated AI: AI fashions analyzing real-time knowledge from digital well being applied sciences (e.g., wearables) are being reviewed for roles in affected person monitoring, endpoint adjudication, and trial administration.

NAMs and the FDA Risk Framework

Under the January 2025 steerage, the FDA makes use of a two-dimensional danger framework:

1. Model Influence Risk: How a lot the AI mannequin’s output influences choices.

2. Decision Consequence Risk: The potential impression of these choices on affected person security or knowledge integrity.

This determines the extent of documentation and disclosure required. High-risk AI fashions should embody detailed submissions on coaching knowledge, mannequin efficiency, and governance — creating stress with commerce secret safety and rising the strategic want for patent protection. The FDA’s evolving stance on AI-based NAMs alerts that AI-enabled platforms will quickly be customary parts of regulatory filings. As such, builders should plan early for IP and knowledge governance points as mentioned beneath in sections B and C. 

Implications for IP Strategy

The FDA’s steerage signifies that when NAMs and AI fashions are employed in scientific or manufacturing decision-making, stakeholders could also be required to offer disclosures about knowledge sources, mannequin coaching procedures, analysis metrics, and upkeep protocols. As these transparency necessities increase, relying solely on commerce secrets and techniques turns into much less sensible, and patent safety or hybrid IP methods more and more vital. Mapping the precise improvements related to NAMs and AI fashions permits stakeholders to obviously establish patentable innovations. The following framework gives a sensible method for mapping key technological developments implied or related to NAMs and AI fashions past the mannequin itself.

Does the mannequin generate or allow clinically actionable info in new methods?

Discovery of recent scientific info can indicate new technique steps (e.g., particular administration routes), formulations, dosages, and therapies of various indications or affected person populations. As mentioned in our prior evaluation of GLP-1 receptor agonists, whereas newly found mechanisms might not themselves be patentable, they usually allow claims round new strategies of remedy, dosage regimens, or formulations. See GLP-1 Receptor Agonists and Patent Strategy: Securing Patent Protection for New Use of Old Drugs. Accordingly, the usage of a NAM may indicate patentable claims directed to strategies of remedy primarily based on new affected person populations, dosages, or formulations. Alternatively, the usage of NAMs may indicate patentable claims directed to workflows for predicting affected person outcomes or monitoring remedy responses. 

Does the mannequin shift how remedy regimens or trial protocols are designed?

The AI mannequin might, for instance, change the controls and design wanted for the scientific trial, comparable to inclusion and exclusion standards, or novel methods of stratifying sufferers by molecular subtypes, genomic, or epigenomic signature. 

In addition, dynamic AI techniques may change the intervals or timing of administering therapeutic brokers. 

Does the mannequin impose new necessities on knowledge enter or integration?

The AI mannequin might require novel multimodal or longitudinal integration by, for instance, combining imaging, omics, and wearable knowledge. The AI mannequin might incorporate epidemiological knowledge to find out affected person clusters, and mixture with complicated biomarker signatures from massive knowledge on molecular signatures, genomic, epigenomic, or multi-omics knowledge obtained from the sufferers. The use of such AI fashions might subsequently indicate patentable workflows directed to info flows that permit prediction of affected person outcomes, stratification of affected person inhabitants, detection, or monitoring of illness improvement. 

Does the AI mannequin or its knowledge construction change how upstream samples are collected or processed?

The AI mannequin might alter affected person pattern workflows, probably resulting in patentable strategies. Changes may embody new biospecimen preservation protocols, modified pattern amount or sort necessities, or added pre-processing steps. These workflow changes can help claims for pattern preparation strategies, automated techniques built-in with AI, or compositions involving specifically processed samples.

In sum, as AI fashions take on roles as soon as reserved for scientific trials or animal research, their affect on medical decision-making, knowledge assortment, and regulatory outcomes calls for a extra nuanced and forward-thinking IP technique. The framework above gives a structured approach to establish improvements tied not solely to the mannequin itself however to its impression on upstream workflows, remedy paradigms, and lifecycle administration. For innovation leaders and IP groups, the above framework gives a scientific approach to future-proof your AI fashions and guarantee strategic safety.

This strategic lens additionally highlights the necessity for equally rigorous knowledge governance approaches, which we now discover in Section C.

Data Governance and Compartmentalization in the NAM Era

As in silico NAMs and AI-enabled platforms turn out to be central to FDA submissions, knowledge governance emerges as a crucial strategic pillar — each for regulatory compliance and IP safety. Sponsors should now plan not just for knowledge high quality and mannequin efficiency, but in addition for the way knowledge disclosures intersect with aggressive benefit.

FDA steerage emphasizes lifecycle transparency: knowledge inputs, coaching datasets, take a look at cohorts, mannequin validation methods, and even future updates might have to be disclosed and monitored. While these disclosures promote regulatory belief, in addition they pose dangers to commerce secrets and techniques and aggressive differentiation — notably the place mannequin efficiency relies upon closely on proprietary datasets or knowledge preparation pipelines. Many life science corporations are conversant in strong knowledge governance in the privateness context, however privateness issues are distinct from managing confidentiality and disclosure when balancing regulatory compliance with IP safety. To deal with these tensions, builders ought to take into account a tiered knowledge governance technique, emphasizing:

1. Modularization and Compartmentalization of Model Components

• By isolating parts of mannequin design (e.g., preprocessing pipeline, mannequin structure, deployment surroundings), corporations can disclose solely the parts related to a given regulatory context. For instance, designing “Virtual Labs” of AI fashions working collectively may assist modularizing totally different features and knowledge units to facilitate a drained knowledge governance system and restrict the required knowledge disclosure. See for instance our earlier dialogue of the rising use of “Virtual Labs” fashioned by a gaggle of AI fashions with distinct features: The Virtual Lab of AI Agents: A New Frontier in Innovation.

2. Decoupling Proprietary Data from Submission Sets

• Rationale: Training on massive inside datasets whereas validating on publicly shared or FDA-approved units might permit compliance with out disclosing delicate uncooked knowledge.
• Example: Model is educated on proprietary multi-omics knowledge however validated towards FDA-endorsed problem datasets for regulatory assessment.

3. Governance-by-Design for Versioning and Traceability

• Rationale: Lifecycle upkeep of AI fashions — together with knowledge drift, re-training, and re-deployment — have to be documented and justified to FDA. A governance structure that logs adjustments, justifies updates, and auto-generates audit trails is more and more indispensable.
• Example: Automated logs displaying when mannequin weights have been up to date as a consequence of new population-level knowledge traits, with reproducibility assured.

These methods not solely align with the FDA’s evolving expectations, but in addition help future IP assertions — e.g., submitting patents round knowledge partitioning, mannequin upkeep instruments, and regulatory integration pipelines. Similarly, because the FDA’s personal use of AI instruments will increase, corporations ought to preserve an open dialogue on how these instruments are getting used and on what knowledge, in order that they will refine their knowledge governance accordingly.

Conclusion: Aligning Innovation, Transparency, and Strategy

AI-based drug improvement and non-animal strategies have been delivered to the regulatory forefront by FDA’s launch of the January 2025 AI steerage, April 2025 NAM roadmap, and additional milestones in inside AI use. In silico fashions have been accepted for FDA regulatory submissions as exemplified by the UVA/Padova simulator. With these advances come each alternatives and obligations. As outlined above, builders ought to fastidiously assessment FDA’s steerage to find out methods for constructing scalable, protectable, and clinically impactful AI platforms. Strategic eager about IP and knowledge governance ought to start on day one. By pondering early and systematically about IP and knowledge governance, stakeholders can place themselves on the frontline of the AI-powered transformation in drug improvement.

Foley is right here that will help you deal with the short- and long-term impacts in the wake of regulatory adjustments. We have the sources that will help you navigate these and different vital authorized issues associated to enterprise operations and industry-specific points. Please attain out to the authors, your Foley relationship accomplice, our Health Care & Life Sciences Sector, or to our Innovative Technology Sector with any questions.

The submit Navigating FDA’s Proposed Guidance on AI and Non-Animal Models: Safeguarding Innovation in Drug Development appeared first on Foley & Lardner LLP.