AI in Ethics Review: Are We Ready?

An ethics committee receives fifty applications each month.

An AI agent screens submissions for completeness, identifies missing documents and sections, checks consistency of consent language against committee requirements, suggests reviewers with relevant expertise, and drafts all correspondence to researchers.

The application reaches reviewers faster. Administrative delays are reduced. The committee spends less time on paperwork and more time discussing ethical issues.

No ethical decision has been made. Yet , this AI agent has already influenced the review process. So where should the line be drawn?

The AI conversation has moved

Much of the discussion around artificial intelligence in research ethics has focused on researchers themselves. The questions commonly asked during review included:

·      How should AI be used in data collection?

·      What are the risks of bias or incomplete data?

·      How should researchers disclose the use of generative AI in their work?

·      Where is this information stored, and how will it be used beyond this project?

These are important questions. But they are no longer the only questions.

Increasingly, ethics committees and research governance bodies are beginning to explore how AI might support the review process itself. Around the world, AI tools are being developed and trialled to assist with application screening, document review, workflow management, and administrative support.

The question is no longer whether AI will become part of research governance. The question is how.

What could AI actually do?

Despite the headlines, the most immediate applications of AI are not about replacing ethics committees. They are about reducing administrative burden.

Many ethics committees and their secretariats spend significant time managing processes that sit around ethical review rather than ethical judgement itself. Applications arrive incomplete. Documents require pre-screening. Reviewers need to be identified and assigned. Post-review, amendments and reports need monitoring.

These tasks are essential, but they are not necessarily the same as deciding whether a research project is ‘ethically acceptable’.

AI is increasingly capable of supporting these functions. It can identify missing information, compare documents against each other for consistency, summarise lengthy applications, and help organise review workflows. Used appropriately (and transparently), these tools could improve efficiency and reduce delays for researchers and committees alike.

Administrative support is not 'Ethical Judgement'?

This distinction is critical. Ethics review is not simply a process of checking boxes. It involves balancing competing interests, assessing risk, considering context, and making judgements about what is fair, proportionate, and respectful.

As an example, a participant information sheet may technically contain all the required elements. An AI agent identify that. But whether the information is genuinely understandable to the intended participants is a different question; a skill that is often honed by reviewers (and the Secretariat) through their own experience and expertise.

Likewise, a project may appear low risk on paper. An AI system can highlight standard risk indicators, no problem. But understanding cultural context, community expectations, power dynamics, or the broader implications of a study requires more than pattern recognition.

These are not merely procedural questions; they are ethical ones. And ethical judgement remains a fundamentally human responsibility.

A system in transition

Aotearoa’s research ethics system is entering a period of significant change.

For many years, operational guidance around ethics committee processes and governance sat under the functions of the Health Research Council Ethics Committee. As those functions transition into a new structure centred around the National Ethics Advisory Committee, there is an opportunity to think more broadly about the future of research governance.

This conversation is not simply about standards or policy. It is also about process and understanding ‘quality’ of ethics review.

Whether you agree with it or not, AI is here to stay. Whilst much of the public discussion focuses on researchers using AI, a quieter conversation is emerging around the use of AI within ethics review systems themselves. As our ethics landscape evolves, it may be worth asking not only what ethical standards we expect committees to uphold, but also what tools and systems might help them do so effectively.

The challenge will be ensuring that any adoption of AI strengthens transparency, consistency, and accountability, rather than simply automating existing processes.

The risk may not be what we think

When discussing AI in ethics review, it is easy to imagine a future where machines replace committees altogether. In reality, the more immediate risk is far more subtle.

An AI system that summarises applications may influence which issues reviewers focus on. Automated screening tools may shape perceptions of risk. Recommendations generated by an AI assistant may carry an authority that exceeds their actual reliability.

This phenomenon, often described as automation bias, occurs when people place too much trust in automated outputs simply because they have been generated by a system.

Therefore, the potential risk is not that AI makes decisions; it is that humans stop questioning them.

Accountability for ethical decisions must remain with people.

Researchers are not seeking approval from a machine. They are seeking assurance that their work has been independently and thoughtfully considered. That expectation should not change.

Are We Ready?

The question is not whether AI has a role in ethics review. It almost certainly does.

The more important question is whether ethics committees are prepared to establish clear boundaries around how these tools are used, what responsibilities remain with human reviewers, and how transparency and accountability will be maintained.

As AI becomes increasingly embedded within research systems, these conversations will become unavoidable. The challenge is ensuring that innovation strengthens ethical review rather than quietly reshaping it without scrutiny.

Final Thoughts

At The Collective, we are closely following the development of AI within research governance and ethics review. We believe there is genuine potential for AI to support administrative efficiency, improve consistency, and reduce barriers for researchers.

However, the values that underpin ethical review — accountability, transparency, independence, and human judgement — cannot be delegated to an algorithm. As research governance continues to evolve, our focus remains on ensuring that innovation strengthens ethical oversight rather than replacing it.