AI Consent Language Guide

Example Consent Language for Artificial Intelligence (AI) Research

When investigators will use AI in research (for example, in recruitment, data collection, analysis, or interventions), they should incorporate the language below (as applicable) into their consent, child assent, and/or parent permission form(s) so that participants may consider the implications of AI usage in the study. This language must be tailored to the specific study.

The examples below are organized by the typical consent sections found in our consent form templates.

Key Information

Briefly state what kind of AI research the participant is joining. Example language:

  • The purpose of this study is to [develop/validate/test] an artificial intelligence (AI) system designed to [state function].

Purpose

Explain the overall purpose of the research, clearly describing how AI is used using plain language. Example language:

  • The purpose of this study is to [develop/validate/test] a new Artificial Intelligence (AI) system that can [describe purpose]. Your data will help improve the AI’s ability to [study goal].
  • This study explores how AI interprets human input. You will interact with an AI model, and your responses will help improve future systems.
  • AI technology, [name/type], will collect [describe data] to help us understand [study goal].
  • The AI system is experimental and may not always be accurate. We will compare its results with human judgement.

Procedures

Describe procedures, specifying how the AI uses data as part of the research activities. Example language:

  • You will interact with an AI system by [describe method]. You will be asked to [tasks].
  • The AI will analyze your [type of data] to [describe purpose].
  • In a simulated environment, the AI will [describe role]. Using [interface], you will [tasks].
  • The AI adapts based on your inputs, so it may respond differently to you than to other participants.
  • We are asking for your permission to use your [type of data] to help train an AI [type].

Benefits

State any direct benefits to participants or future benefits to others, society, or with future AI tools. Example language:

  • The benefits of taking part in this research may be [list possible direct benefits]. – OR – There will be no direct benefit to you from taking part in this AI research.
  • It is hoped that the information gained from the study will help improve the [accuracy, safety, and fairness] of future AI tools that support [describe purpose].

Risks/Discomforts

List potential AI-related risks, how they will be mitigated. Include information about participants’ right to stop participation. Example language:

  • In studies involving AI, the main risk is to your privacy. Even when your name and contact information (or other identifiers) are removed, AI systems can detect complex patterns that, in rare cases, could make re-identification possible.
  • Interactions with AI may feel [e.g., repetitive or frustrating]. We will minimize this by [mitigation].
  • The AI is still in development and may generate recommendations that produce errors or unexpected results. There is a small risk that AI-generated recommendations may be incorrect or misleading. [Modify these statements for project specifics: Researchers and clinicians are trained to review all AI outputs carefully and are required to question or override recommendations that appear inaccurate. The study team monitors AI performance and takes steps to correct errors.]
  • Rarely, harmful or offensive AI-generated content may appear. If this occurs, [actions participants should take]. Support will be available.
  • You may stop participating or request a non-AI alternative at any time.

Confidentiality

Specify whether AI tools are UCB-licensed commercial tools, external commercial tools, or institutional. Outline confidentiality safeguards including whether data will be deidentified. Explain how data will be managed after study completion including de-identification, secondary use, sharing plans, and data retention time frames. State clearly whether data will be used to train commercial AI software, whether AI tools or vendors will retain data and how long. Acknowledge withdrawal limits and unknown future risks. If applicable, explain that this AI research may collect and produce data or other techniques that can lead to commercial profit. State whether results generated by the AI will be shared with the participant, particularly if the AI detects unexpected health/behavioral issues. Example language:

  • We will use [AI tool(s)] to [describe use]. [AI tool] is [a UCB-licensed system/an external commercial system/a UCB institutional system]. [AI company name] [will/will not] retain [identifiable/de-identified] data.
  • The AI tool will access only deidentified data. – OR – The AI tool will access [specific identifiers] to [reason].
  • Your data [will/will not] be used to train the AI system. [If will: Once used to train AI, data typically cannot be withdrawn. We may continue using your de-identified data to improve AI systems or share it in research repositories. Future techniques could slightly increase re-identification risk.]
  • Data processed by the AI will be securely deleted within [timeframe]. Your data [can/cannot] be withdrawn after being entered into the AI system.
  • This research will use a third-party transcription service, [AI tool name] that uses Artificial Intelligence (AI) to transcribe recordings into text. The service [may use/will not use] your recordings for training their AI models.
  • It is possible that the AI tool developed in this study may one day become part of a commercial product (for example, a software tool that [purpose]). Should this occur, you will not share in any commercial value or profit derived from the use of your data and/or information obtained from them.
  • [Individual/group] results generated by the AI [will/will not] be shared with you. [If individual results will be shared: If we find any information about your health that your regular doctor needs to know about (because it could impact your medical care), we will [describe how and to whom].]

Related information