Opens in a new tab

Home / Resources / AI in Research

AI in Clinical Research:
What FDA, EMA and ICH E6(R3) Expect

Regulatory guidance on AI in clinical research is evolving, and the main documents have different purposes and legal status. FDA material addresses particular regulatory decision-making uses of AI; EMA guidance covers the medicinal-product lifecycle; ICH E6(R3) provides broader good clinical practice principles. None of these should be paraphrased as a universal ban on AI-assisted statistical programming (FDA, 2025; EMA, 2024, 2025).

AI in Research
WHO ASKSSponsors, regulators
WHEREFDA, EMA, ICH
KEEPValidation evidence
NEVERCall a draft final
LAST CHECKEDNovember 2026

This is a guide to published regulatory documents for research planning, not legal advice or a substitute for the current sponsor and regulator requirements.

Which regulatory guidance on AI in clinical research applies?

POLICY TABLE · Last checked November 2026 · rechecked every six months

Source Scope Status or distinction as of November 2026
FDA January 2025 AI document AI-produced information intended to support regulatory decisions on safety, effectiveness or quality of drug/biologic products Draft guidance: recommendations, not final binding rules
EMA AI reflection paper (2024) AI use across medicinal product development, authorization and postauthorization Reflection paper expressing regulatory considerations
FDA-EMA good AI practice principles (January 2026) Broad principles for trustworthy AI across drug development Joint guiding principles; not a standalone regulation
ICH E6(R3) Principles and Annex 1 Good clinical practice, trial conduct and data quality Effective in the EU on EMA’s published timeline since July 2025
ICH E6(R3) Annex 2 Trials with additional design features and data sources Adopted in 2026; EU effective date January 15, 2027 (EMA, 2026)

The implementation dates above reflect EMA’s published status; jurisdiction-specific adoption and sponsor obligations should be checked separately.

What does the FDA January 2025 draft actually address?

The FDA document, Considerations for the Use of Artificial Intelligence to Support Regulatory Decision-Making for Drug and Biological Products, concerns the credibility of AI models used to generate information supporting regulatory decisions. It adopts a risk- and context-of-use perspective (FDA, 2025).

It should not be presented as a final rule covering every researcher who uses an LLM to draft generic R syntax. A high-stakes AI model supporting a drug-submission claim and a chatbot suggesting a code comment pose different risks and require different documentation.

What is the EMA's position?

The EMA’s 2024 reflection paper considers AI across the medicinal-product lifecycle, including data reliability, model performance, governance and human oversight (EMA, 2024). In January 2026, FDA and EMA jointly presented 10 guiding principles of good AI practice in drug development, emphasizing a risk-based approach and reliable, transparent use (FDA & EMA, 2026).

The practical message is that AI use can be acceptable within a controlled evidence-generation process. The necessary validation depends on the particular model, purpose, input data and consequences of error.

Where does ICH E6(R3) fit?

ICH E6(R3) focuses on good clinical practice, quality by design, proportionate risk management, reliable records and suitable systems. Its scope is broader than AI. Applying its principles to AI-supported data processing involves showing that the system and workflow are fit for the intended purpose and that critical data remain trustworthy (EMA, 2025).

For example, a protocol-specific analysis script should be tested against the SAP and validated outputs. Where external AI processing of regulated trial data is proposed, the sponsor should assess vendor qualification, contracts, information security, data integrity and traceability in the actual configuration, not just the tool’s brand name.

How to organize the validation argument

  1. Define the intended use. Is AI drafting generic code, extracting evidence, predicting an endpoint, or generating information used in regulatory decisions?
  2. Identify the data. Are records synthetic, public, confidential, pseudonymized or patient-level?
  3. Describe the system. State versions, roles, access controls and relevant changes.
  4. Match controls to risk. Include performance testing, independent programming checks, traceability and governance as appropriate.
  5. Align analysis with trial objectives. Verify endpoint definitions, estimands, SAP and sensitivity analyses.
  6. Maintain evidence. Record what was tested, by whom, and what changed before the result was accepted.

This is an analytical framework drawn from the cited guidance rather than a prescribed set of six legal requirements. Independent Statistical Review covers analyses prepared with AI assistance.

What about peer-reviewed AI reporting guidelines?

Reporting guidelines such as CONSORT-AI and TRIPOD+AI describe what authors should report in specific types of studies; they are not the same as regulatory authorization. The appropriate guideline depends on whether AI is the intervention being evaluated, a clinical prediction model, or only an assistant for programming (Liu et al., 2020; Collins et al., 2024).

Frequently asked questions

The cited FDA document does not establish such a universal prohibition. Actual use must satisfy the applicable study, sponsor, data-protection and regulatory requirements.

No. It is a broad good clinical practice guideline. Its data integrity and system-quality principles can be relevant to an AI-assisted workflow.

The document checked for this article was labeled draft guidance. Check the FDA publication page again before publishing or applying it to a live submission.

Independent Statistical ReviewNeed a statistician to confirm the analysis before you submit? An in-house PhD-level statistician reviews the analysis against your protocol and SAP, while sponsor and regulator requirements stay with your team.

Request an Assessment

Independent Statistical Review

OnlineSPSS is a specialist statistical consultancy for researchers, and Independent Statistical Review is carried out by in-house PhD-level statisticians. The review checks method, assumptions, reproducibility and reporting. Overall responsibility for the study remains with the researcher or research team.

References

Last reviewed November 2026 by the OnlineSPSS Statistical Team. Worked examples use hypothetical values and are not client data.

NEXT STEP

Need this done for
your study?

Share your research in about a minute and receive a complimentary assessment the same day: scope, timeline and one fixed fee.

No Commitment Required

Strictly Confidential

Fixed Project Fee

In-House PhD-Level Statisticians

Same-Day Response

Serving Researchers Since 2014