Does Using Artificial Intelligence (AI) Violate IRB Requirements?

Does Using Artificial Intelligence (AI) Violate IRB Requirements?

Artificial intelligence tools such as ChatGPT, Gemini, and Claude are increasingly used in doctoral research for writing, coding, analysis support, and brainstorming. A common concern among researchers is: Does using AI violate IRB requirements?

The short answer is: not necessarily — but it can, depending on how AI is used, what data is involved, and what your approved protocol allows. IRB (Institutional Review Board) requirements are not about AI specifically. They are about protecting human participants, ensuring ethical data handling, and maintaining research integrity. AI use becomes an IRB issue only when it affects those principles.

What IRB Requirements Actually Focus On

IRB approval is primarily concerned with:

• Informed consent

• Participant privacy and confidentiality

• Data security and storage

• Risk to participants

• Ethical handling of human subject data

• Compliance with institutional and legal standards (e.g., GDPR in Europe, HIPAA in the USA)

Nowhere in most IRB frameworks is AI explicitly banned. However, AI tools can intersect with these requirements in important ways.

When Using AI DOES NOT Violate IRB Requirements

In many cases, AI use is completely acceptable.

1. Writing and Editing Support

Using AI tools like ChatGPT for:

• Grammar correction

• Improving clarity

• Editing academic writing

• Structuring paragraphs

is generally not an IRB issue because:

• No participant data is involved

• No confidentiality risk is introduced

• It does not affect data collection or consent

2. Methodology Planning and Brainstorming

AI may be used to:

• Generate research ideas

• Help design interview questions

• Suggest analysis approaches

• Outline dissertation chapters

This is also typically outside IRB scope because it does not involve human subject data.

3. Statistical or Coding Assistance (Without Data Upload)

Using AI to:

• Write R or Python code

• Explain SPSS output

• Debug scripts

is usually acceptable as long as:

• No identifiable data is uploaded

• The approved data handling protocol is followed

When AI CAN Violate IRB Requirements

AI becomes an IRB issue when it involves human participant data or violates approved procedures.

1. Uploading Identifiable Participant Data

Uploading to AI tools like ChatGPT:

• Interview transcripts

• Survey responses

• Audio transcripts

• Any indirect identifiers

may violate IRB requirements if:

• It was not explicitly approved

• Data leaves secure institutional systems

• Confidentiality agreements are breached

2. Using AI Outside Approved Data Storage Procedures

IRB protocols often specify:

• Where data must be stored

• Who can access it

• Whether third-party systems are allowed

Using external AI tools without approval may violate these conditions.

3. Changing Data Handling Without Amendment Approval

If you:

• Decide mid-study to use AI for analysis

• Upload data not originally approved for external processing

you may need an IRB amendment. Otherwise, it can be a compliance violation.

4. Risking Re-Identification of Participants

Even if data is anonymized, AI processing can:

• Combine contextual clues

• Reconstruct sensitive patterns

• Increase risk of identification

This can conflict with IRB confidentiality assurances.

Why IRB Boards Are Concerned About AI

IRB concerns are not about the technology itself, but about:

1. Data Security

AI tools may process data on external servers.

2. Lack of Control Over Data Flow

Researchers may not fully know how data is stored or processed.

3. Compliance With Privacy Laws

Regulations like GDPR require strict control over personal data.

4. Transparency and Accountability

IRB wants clarity on how data is handled throughout the study.

Does AI Automatically Require IRB Approval?

Not always.

You generally do NOT need IRB approval to use AI for:

• Writing assistance

• Literature reviews

• Idea generation

• Coding help without data input

However, you MAY need approval if:

• AI processes raw participant data

• AI is part of your data analysis pipeline

• AI influences interpretation of human subject data

Some institutions now require disclosure of AI use in research methods or appendices.

Best Practices for Staying IRB-Compliant When Using AI

1. Follow Your Approved Protocol Exactly

Your IRB application is your binding agreement.

2. Avoid Uploading Raw Participant Data

Do not upload:

• Interviews

• Open-ended survey responses

• Sensitive datasets

into tools like ChatGPT unless explicitly approved.

3. Use AI for Indirect Support Only

Safe uses include:

• Writing support

• Coding assistance without data

• Concept clarification

• Statistical explanation

4. Request an IRB Amendment if Needed

If you want to:

• Introduce AI into data analysis

• Change data processing workflows

you may need formal approval.

5. Check Institutional AI Policies

Many universities now have separate guidance on:

• Generative AI use

• Data privacy requirements

• Approved tools and platforms

Example Scenarios

Safe (No IRB Issue):

Using ChatGPT to:

• Improve dissertation wording

• Generate interview questions

• Help debug Python code

Potential IRB Violation:

Uploading full interview transcripts into AI tools for thematic analysis without approval.

Borderline Case:

Using AI to analyze anonymized excerpts without knowing whether indirect identifiers remain — may require IRB clarification.

Final Thoughts on Does Using Artificial Intelligence (AI) Violate IRB Requirements?

Using AI does not automatically violate IRB requirements, but it becomes a compliance issue when it involves:

• Human subject data

• Privacy or confidentiality risks

• Deviations from approved protocols

• Unapproved external data processing

AI tools like ChatGPT can be valuable research assistants, but IRB compliance depends entirely on how they are used, not whether they are used.

In doctoral research, the guiding principle is simple:

If AI interacts with participant data in any meaningful way, check your IRB approval first.

When in doubt, the safest and most professional step is always to consult your dissertation chair or IRB office before proceeding.

Need help with your IRB application? Visit our website.