IRB And Ethics In AI Research: What Changes, What Doesn't
Using AI in human-subjects research raises new IRB questions. Here's how to get approved without surprising your review board.
40 min · Reviewed 2026
The classic IRB checklist still applies
Informed consent, properly documented
Risk minimization and justification
Data security plan
Equitable participant selection
Privacy protections commensurate with sensitivity
What AI adds to the checklist
If AI analyzes participant data, the consent form must say so
If data is sent to a third-party API (OpenAI, Anthropic, Google), the consent AND the data security plan must disclose this — and often the vendor's data-use policy must be attached
If the AI could re-identify anonymized data, the IRB needs to know
If the AI generates content that participants will see (chatbot-based research), the IRB needs the exact model version and prompt frozen
If AI scoring affects participant outcomes, bias testing is typically required
The ethics of LLM participants
Some researchers now use LLMs as 'synthetic participants' — simulating survey responses or interview answers. This is cheap and fast. It is also a minefield: LLMs carry training-data biases, and findings from synthetic participants may not generalize. IRBs increasingly ask whether synthetic-participant studies count as 'human subjects research' at all. Answer this for your project before you submit.
The big idea: AI doesn't create new ethical categories — but it intensifies old ones. Disclose specifically, protect participants aggressively, and talk to your IRB before you submit.
IRB Protocol Amendments: AI-Assisted Drafting for the Iterative Reality of Research
The premise
Amendment drafting is a recurring administrative burden; AI accelerates it without replacing the substantive risk-benefit analysis.
What AI does well here
Draft amendment narratives for common scenarios (new study site, new investigator, recruitment material change)
Generate revised consent forms with tracked changes and a brief change rationale
Produce the amendment justification matching the institution's IRB form structure
Generate the cover letter summarizing the amendment's significance to ongoing risk-benefit analysis
What AI cannot do
Make the regulatory determination of whether an amendment is minor vs. significant
Substitute for the IRB's own substantive review
Generate amendments to consent that haven't been discussed with the study team
Research Ethics Protocols for AI-Assisted Studies
The premise
AI use in research carries new ethics considerations IRBs are now examining; explicit protocol planning prevents review delays.
What AI does well here
Disclose AI use in IRB protocols (analysis, recruitment, intervention, interaction)
Document AI risks specific to your protocol (re-identification, hallucination harm, bias)
Address AI in informed consent when participants interact with AI
Stay current on your institution's AI-research-ethics guidance
What AI cannot do
Substitute AI use disclosure for substantive ethics review
Replace researcher accountability for participant welfare
Anticipate every AI ethics issue (the field evolves)
Using AI to Draft IRB Protocol Deviation Reports
The premise
AI can structure deviation reports covering what happened, impact, and corrective action.
What AI does well here
Describe deviation factually
Propose corrective actions
What AI cannot do
Decide reportability threshold
Speak for the IRB
Understanding "Using AI to Draft IRB Protocol Deviation Reports" in practice: AI is transforming how professionals approach this domain — speed, precision, and capability all increase with the right tools. Document protocol deviations clearly for IRB review — and knowing how to apply this gives you a concrete advantage.
Apply IRB in your research workflow to get better results
Apply deviation in your research workflow to get better results
Apply reporting in your research workflow to get better results
Apply Using AI to Draft IRB Protocol Deviation Reports in a live project this week
Write a short summary of what you'd do differently after learning this
Share one insight with a colleague
AI and an IRB protocol summary rewrite
The premise
IRB lay summaries default to jargon. AI can lower reading level while preserving the regulatory content.
What AI does well here
Lower reading level to 8th grade.
Replace jargon with everyday words.
Preserve risk and benefit language.
What AI cannot do
Decide which risks must be disclosed.
Replace the PI's responsibility for accuracy.
Capture local IRB preferences without you sharing them.
AI and IRB Amendment Summary: Tracking Protocol Changes
The premise
AI can diff two protocol versions and produce an IRB amendment summary that organizes changes by IRB-relevant category.
What AI does well here
Categorize changes (consent, eligibility, procedures, risk) per IRB norms
Surface changes that may trigger re-consent of enrolled participants
What AI cannot do
Decide whether a change is minor or substantive under the IRB's policy
Sign the IRB submission as the PI
AI and IRB Protocol Skeletons: Human-Subjects Drafting
The premise
AI can take a study summary and produce an IRB protocol skeleton with sections for purpose, recruitment, consent, risks, and data.
What AI does well here
Produce the standard IRB section structure
Suggest standard consent language as a starting point
What AI cannot do
Make ethical judgments about risk and benefit
Replace IRB review or institutional sign-off
End-of-lesson check
15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-research-irb-ethics-creators
What is the core idea behind "IRB And Ethics In AI Research: What Changes, What Doesn't"?
Using AI in human-subjects research raises new IRB questions. Here's how to get approved without surprising your review board.
counter-argument
'Give me a kid example for the word 'gravity.''
Cost, ethics, feasibility, and novelty are all human calls
Which term best describes a foundational idea in "IRB And Ethics In AI Research: What Changes, What Doesn't"?
informed consent
IRB
HIPAA
BAA
A learner studying IRB And Ethics In AI Research: What Changes, What Doesn't would need to understand which concept?
IRB
HIPAA
informed consent
BAA
Which of these is directly relevant to IRB And Ethics In AI Research: What Changes, What Doesn't?
IRB
informed consent
BAA
HIPAA
Which of the following is a key point about IRB And Ethics In AI Research: What Changes, What Doesn't?
Informed consent, properly documented
Risk minimization and justification
Data security plan
Equitable participant selection
Which of these does NOT belong in a discussion of IRB And Ethics In AI Research: What Changes, What Doesn't?
Data security plan
counter-argument
Risk minimization and justification
Informed consent, properly documented
Which statement is accurate regarding IRB And Ethics In AI Research: What Changes, What Doesn't?
If data is sent to a third-party API (OpenAI, Anthropic, Google), the consent AND the data security …
If the AI could re-identify anonymized data, the IRB needs to know
If AI analyzes participant data, the consent form must say so
If the AI generates content that participants will see (chatbot-based research), the IRB needs the e…
Which of these does NOT belong in a discussion of IRB And Ethics In AI Research: What Changes, What Doesn't?
If AI analyzes participant data, the consent form must say so
counter-argument
If data is sent to a third-party API (OpenAI, Anthropic, Google), the consent AND the data security …
If the AI could re-identify anonymized data, the IRB needs to know
What is the key insight about "HIPAA and APIs" in the context of IRB And Ethics In AI Research: What Changes, What Doesn't?
Sending PHI to a third-party LLM without a Business Associate Agreement is almost always a HIPAA violation.
counter-argument
'Give me a kid example for the word 'gravity.''
Cost, ethics, feasibility, and novelty are all human calls
What is the key insight about "Disclose early" in the context of IRB And Ethics In AI Research: What Changes, What Doesn't?
counter-argument
IRB coordinators HATE being surprised. If your protocol uses AI in any non-trivial way, email the IRB office before you …
'Give me a kid example for the word 'gravity.''
Cost, ethics, feasibility, and novelty are all human calls
What is the key warning about "Maintain methodological rigour" in the context of IRB And Ethics In AI Research: What Changes, What Doesn't?
counter-argument
'Give me a kid example for the word 'gravity.''
AI-assisted research requires transparent disclosure of tools used, validation of outputs against primary sources, and p…
Cost, ethics, feasibility, and novelty are all human calls
Which statement accurately describes an aspect of IRB And Ethics In AI Research: What Changes, What Doesn't?
counter-argument
'Give me a kid example for the word 'gravity.''
Cost, ethics, feasibility, and novelty are all human calls
Some researchers now use LLMs as 'synthetic participants' — simulating survey responses or interview answers. This is cheap and fast.
What does working with IRB And Ethics In AI Research: What Changes, What Doesn't typically involve?
The big idea: AI doesn't create new ethical categories — but it intensifies old ones.
counter-argument
'Give me a kid example for the word 'gravity.''
Cost, ethics, feasibility, and novelty are all human calls
Which best describes the scope of "IRB And Ethics In AI Research: What Changes, What Doesn't"?
It is unrelated to research workflows
It focuses on Using AI in human-subjects research raises new IRB questions. Here's how to get approved without sur
It applies only to the opposite beginner tier
It was deprecated in 2024 and no longer relevant
Which section heading best belongs in a lesson about IRB And Ethics In AI Research: What Changes, What Doesn't?
counter-argument
'Give me a kid example for the word 'gravity.''
What AI adds to the checklist
Cost, ethics, feasibility, and novelty are all human calls
