What Should Researchers Do When AI Is Introduced After Participants Have Already Consented?

What if Your Participants Already Consented Before You Decided to Use AI?

You are conducting a qualitative study involving university instructors. Participants have already signed consent forms and completed interviews about their experiences with institutional policies concerning generative AI.

At the time of recruitment, the study involved researcher-led thematic analysis. The participant information sheet explained that transcripts would be examined by the research team and stored according to the university's approved data management procedures.

Several weeks later, you discover an AI service that could help organize transcripts and generate preliminary coding suggestions.

You believe the tool could make analysis more manageable, but the original consent materials did not mention generative AI or an external processing provider.

Should you contact every participant again? Can you rely on the original consent because the research questions have not changed? Or must the institution first determine whether the proposed AI processing is consistent with the conditions under which participants agreed to the study?

Review the Original Consent and Obtain the Appropriate Determination Before Processing

How to Introduce AI After Participants Have Already Agreed to a Study

Consent Applies to the Research Participants Were Told About

Research participation consent is obtained on the basis of information provided about the proposed study.

That information may describe the purpose, procedures, risks, confidentiality arrangements, and relevant uses of participant information.

For example, participants may agree to interviews that will be analyzed by the university research team using institutionally managed systems.

Introducing an external AI provider may create a processing arrangement that differs from what participants were originally told.

However, not every software change necessarily alters the substance of the consent arrangement.

Researchers should examine the original materials and the proposed activity rather than assume that consent either covers every future analytical method or becomes invalid whenever a new tool is introduced.

First Identify Exactly What Has Changed

The appropriate response depends on the difference between the original procedure and the proposed AI-assisted workflow.

For example, using AI to improve the grammar of a nonconfidential manuscript may have little relevance to participant information.

By contrast, submitting identifiable interview transcripts to an external AI service may introduce a new data recipient, retention arrangement, or confidentiality risk.

These are illustrative considerations rather than automatic classifications of which changes require reconsent.

The relevant institution should determine the appropriate procedure under its governing requirements.

Distinguish an Ethics Amendment From Participant Reconsent

Researchers sometimes assume that any change requiring ethics review must also require every participant to sign a new consent form.

However, these are different decisions.

An institution may determine that a proposed AI-related modification requires review but does not require renewed consent.

Alternatively, a consequential change may require both an amendment and additional participant information or agreement.

The decisions should not be treated as interchangeable.

Researchers should follow the applicable procedures for introducing AI after research ethics approval while separately evaluating the consent implications.

Does the Original Consent Form Mention Data Analysis?

Many consent forms explain that participant responses will be used for research analysis without identifying every analytical technique.

Whether this language covers AI-assisted processing depends on the scope of the consent, the actual processing arrangement, and applicable requirements.

For example, an authorized analytical tool used within the original confidentiality conditions may raise different questions from a new external provider receiving identifiable transcripts.

Researchers should not interpret a general reference to data analysis as unrestricted permission for every future form of processing.

However, they should also avoid assuming that the absence of the word "AI" automatically invalidates the original consent.

The relevant issue is whether the proposed change materially alters the research procedures or information participants were given.

Why the Original Confidentiality Statement Matters

Participants may have agreed to the study partly because of assurances concerning who would access their information.

For example, a consent form may state that interview transcripts will be accessible only to authorized research team members.

If the researcher later proposes submitting those transcripts to an external AI provider, the processing arrangement may differ from the original explanation.

However, external processing does not automatically mean the information becomes public or that confidentiality is necessarily breached.

The researcher should examine whether the provider's role and safeguards are consistent with the approved arrangements.

The question of whether an AI provider becomes a recipient of research data may be central to the assessment.

Can De-Identification Make Reconsent Unnecessary?

Removing direct identifiers may reduce privacy risks, but it does not automatically establish that additional consent or institutional review is unnecessary.

For example, a transcript may remain identifiable through distinctive institutional positions, professional experiences, or contextual details.

Even where information is genuinely anonymous, institutional research agreements or the approved protocol may impose restrictions on its use.

Researchers should therefore examine the actual identifiability, processing purpose, and authorization.

De-identification is a safeguard, not an automatic substitute for evaluating the consent implications of a new workflow.

What If the New AI Use Involves Sensitive Inferences?

Introducing AI may expand the analytical activity beyond the original research purpose.

For example, a study may have collected interviews about instructors' teaching practices, while the researcher later proposes using AI to infer participants' psychological characteristics.

Such inferences may be unsupported, unnecessary, or sensitive.

The fact that the original transcripts were collected legitimately does not automatically authorize every characteristic that could be inferred from them.

Researchers should examine whether the new purpose requires additional ethical assessment, participant information, or authorization.

The concern is related to AI-generated sensitive inferences and the limits of the original research purpose.

When Might Participants Need to Be Contacted Again?

Additional participant contact may be appropriate or required when a proposed change materially affects the information on which participants based their decision to participate.

For example, an external AI processing arrangement may introduce a new recipient or retention condition that was not covered by the original consent materials.

However, the appropriate response may involve updated participant information, an additional consent process, or another institutionally authorized procedure.

Researchers should not assume that every change requires renewed signatures from all participants.

Equally, they should not assume that participants never need to be contacted because data collection has already ended.

The specific question of whether reconsent is required when AI changes data processing should be resolved under the applicable framework.

What If Some Participants Cannot Be Contacted?

Researchers may be unable to contact participants who have completed the study, changed contact information, or withdrawn from further communication.

This practical difficulty does not automatically authorize a new AI processing arrangement.

For example, a researcher should not assume that previously collected transcripts may be submitted to an external AI provider simply because obtaining renewed consent would be inconvenient.

Depending on the governing framework, an institution may consider whether additional consent is required, whether a permissible waiver or alternative procedure is available, or whether the proposed processing should not proceed.

Researchers should not independently declare a waiver or infer consent from the absence of a response.

Where necessary authorization cannot be established, the original approved analytical procedure may remain the appropriate alternative.

What If Participants Decline the New AI Processing?

If additional participant agreement is required and a participant declines, the research team should respect the applicable consent and processing arrangements.

For example, a participant may agree to continued researcher-led analysis but refuse submission of their transcript to an external AI service.

Whether that alternative is available depends on the approved study design and institutional requirements.

Researchers should not assume that refusal of AI processing automatically means the participant withdraws from every aspect of the study.

However, they should not promise continued participation through an alternative workflow unless it is feasible and authorized.

The relevant options are discussed in participant opt-out from AI processing.

What If AI Has Already Been Used Before the Researcher Checks?

Sometimes researchers may submit participant information to an AI service before recognizing that the processing differs from the original consent or approved protocol.

For example, a researcher may upload interview transcripts to an external chatbot and later discover that the service was not authorized for those data.

In such circumstances, the researcher should stop further unauthorized processing and promptly consult the appropriate institutional ethics and data-protection personnel.

The institution may need to assess whether a protocol deviation, privacy incident, or other reportable event occurred.

Deleting the conversation does not necessarily reverse the earlier transfer or establish that all information has been erased.

Researchers should preserve relevant factual records and follow institutional reporting procedures rather than attempt to conceal or retroactively characterize the activity as approved.

How Should Researchers Review the Original Consent Materials?

A practical review should compare the original participant information with the proposed AI-assisted workflow.

The researcher should identify what participants were told about data analysis, confidentiality, external recipients, retention, and the purposes of processing.

Next, the researcher should determine which aspects of the proposed AI activity differ from those descriptions.

Finally, the research team should seek the appropriate institutional determination about whether additional participant information or agreement is necessary.

The review should be documented sufficiently to explain why the proposed processing was considered covered, modified, or rejected.

How Should Researchers Explain a New AI Procedure to Participants?

Where additional participant information is required, the explanation should identify what has changed and how the change affects the use of their information.

This example describes a proposed change rather than an approved one.

Researchers should revise the wording to reflect the actual provider, processing conditions, institutional determination, and participant choices.

It should not be used as a substitute for a formal reconsent process where one is required.

Can AI Help Compare the Original and Proposed Consent Arrangements?

Generative AI can assist researchers in identifying differences between the original protocol and proposed AI processing.

However, it should not independently determine whether the original consent legally authorizes the new activity.

Researchers should use nonconfidential descriptions or an appropriately authorized environment when seeking AI assistance with the review.

The final determination should follow the applicable institutional process.

Does the GDPR Require Reconsent Whenever AI Is Introduced?

No universal GDPR rule requires renewed consent merely because a new AI tool is introduced.

The relevant obligations depend on the lawful basis for processing, the original purposes, the proposed further processing, transparency requirements, and applicable data subject rights.

For example, a change in processing purpose may raise different questions from a technical substitution that remains within an existing authorized arrangement.

Where consent is the lawful basis, the scope and validity of that consent must be examined.

Where another lawful basis applies, the researcher must evaluate the relevant requirements rather than assume that research participation consent and data-processing consent are identical.

Researchers should also consider the GDPR's provisions concerning purpose limitation and further processing where applicable.

What Does the Philippine Data Privacy Act Mean for Post-Consent AI Use?

In the Philippines, the Data Privacy Act of 2012 establishes requirements concerning transparency, legitimate purpose, proportionality, and lawful personal information processing.

Introducing an external AI provider may change relevant processing arrangements, including the organizations involved or the purposes for which information is used.

Researchers should assess whether the original participant information and applicable legal basis cover the proposed activity.

However, the Act should not be interpreted as requiring every software modification to trigger a new research consent form.

The appropriate procedure depends on the actual processing and applicable National Privacy Commission and institutional requirements.

Researchers should consult institutional data protection and research ethics personnel when the proposed change may affect participant information or rights.

Can the Original Research Continue While the AI Change Is Being Reviewed?

Potentially. Researchers may be able to continue procedures already authorized under the original protocol while awaiting a determination about the proposed AI-assisted modification.

For example, the research team may continue researcher-led thematic analysis using approved institutional systems.

However, where prior approval is required, the proposed new AI processing should not begin merely because an amendment request has been submitted.

Researchers should follow the institution's instructions concerning which activities may continue and which must await authorization.

The original approval does not necessarily authorize every subsequently proposed method of processing participant information.

What If the Institution Determines That Reconsent Is Unnecessary?

An institution may determine that the proposed AI activity is covered by existing arrangements or that additional participant consent is not required under the applicable framework.

Researchers should follow the determination and any conditions attached to it.

However, a finding that reconsent is unnecessary does not automatically establish that the selected AI service satisfies every data-protection or information security requirement.

Separate institutional authorizations may still apply.

Researchers should document the relevant determination and implement only the processing arrangement actually permitted.

Why Participant Expectations Still Matter

Even when a proposed AI processing arrangement is technically permitted, participants may have understood the original research procedures differently.

For example, a participant may have expected their interview to be examined exclusively by the university research team.

Learning that an external AI service will process the transcript may raise questions about confidentiality and the scope of the original agreement.

This does not automatically establish that the original consent was invalid or that the new processing is prohibited.

However, it provides a reason to examine whether the participant information was sufficiently clear and whether additional explanation would support ethical transparency.

The issue is explored in AI use and research participant expectations.

When Is the Appropriate Decision Not to Introduce AI?

Sometimes the proposed AI workflow may not be compatible with the original consent arrangements, and obtaining additional authorization may be impractical or inappropriate.

For example, the transcripts may contain highly sensitive information that cannot be sufficiently protected through the available external service.

Alternatively, the expected analytical benefit may be limited compared with the additional processing risks.

In such circumstances, the researcher may retain the original approved procedure rather than introduce AI.

The objective is not to force a technological change into an existing study.

Researchers should recognize when not using AI is the more defensible ethical decision.

Introducing AI After Interview Participants Have Signed Consent Forms

The research questions remain unchanged, but the proposed processing arrangement may differ from what participants originally understood.

The appropriate response is to assess the actual modification rather than assume that either automatic reconsent or unrestricted AI use is required.

Common Misconceptions About Introducing AI After Research Consent

Compare the New AI Workflow With What Participants Originally Agreed To

Before introducing AI into a study with existing participants, review the original consent materials and approved protocol.

Identify the proposed changes to information processing, confidentiality, analytical purpose, and potential risks. Then obtain the appropriate institutional determination about whether additional participant information or consent is necessary.

Document the institutional determination, any updated participant information, and the final authorized workflow.

The researcher should not treat earlier consent as an unlimited permission to introduce new processing arrangements, but neither should every technological change be assumed to invalidate the original agreement.

Before Introducing AI After Participants Have Consented

Frequently Asked Questions About Introducing AI After Research Consent

Earlier Consent Does Not Automatically Cover Every Later AI Processing Decision

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.