2RICHARDS PERSPECTIVES · FEATURED PUBLICATION
Governing AI in GCP Clinical Development
Richard Reeve
2Richards
September 2026
Artificial intelligence is beginning to influence how clinical-development professionals identify risks, interpret information and form judgements. This paper examines how established GCP principles can govern that influence without confusing human presence with meaningful oversight.
Its central argument is that AI should earn influence within regulated clinical development. Where it materially affects important decisions, accountable professionals must be able to challenge its contribution, verify the evidence and justify the final judgement.
WHY THIS PAPER
AI does not change the purpose of GCP.
It changes where influence can enter the decision process.
Clinical research depends on evidence that deserves to be trusted. That trust is built through the quality of the decisions, controls and professional judgement applied throughout a trial.
Artificial intelligence introduces a new source of influence into that familiar system. It can shape what professionals notice, how problems are framed and which recommendations receive attention, even where formal responsibility remains unchanged.
The governance challenge is therefore not simply whether AI functions correctly. It is whether sponsors can continue to justify confidence in important decisions when AI begins to influence the reasoning behind them.
MEANINGFUL HUMAN CHALLENGE
The familiar expression “human in the loop” confirms only that a person was present somewhere in the workflow. It does not demonstrate that an AI-generated recommendation was independently evaluated, that assumptions were questioned or that authoritative evidence was consulted.
Meaningful Human Challenge is the disciplined application of professional judgement to an AI-generated contribution before it is permitted to influence an important decision.
Challenge does not mean disagreement. A recommendation may withstand scrutiny and be accepted substantially as presented. The question is whether the degree of reliance was justified by the evidence, context and significance of the decision.
Human presence is not the same as human oversight.
The governance question is not whether a human touched the output. It is whether accountable people subjected it to Meaningful Human Challenge.
WHAT THIS PAPER EXAMINES
Evidence and traceability
What records are needed to demonstrate trustworthy judgement rather than simply document that AI was used.
Sponsor oversight
How established GCP responsibilities apply when AI-supported activities are outsourced or vendor-enabled.
Inspection readiness
Whether the organization can explain where AI influenced important work and why the resulting decisions remained justified.
Credibility before adoption
Whether the AI-enabled function has a bounded purpose and evidence of sufficient performance.
Meaningful Human Challenge
Whether accountable people can genuinely evaluate and challenge AI-generated contributions.
Material influence
When AI begins to affect what is considered, how issues are framed, or why a decision is reached.