37 - Can AI Improve Informed Consent? Opportunities and Risks

Background: : The use of AI platforms in developing informed consent documents raises ethical and regulatory concerns for IRBs. This analysis explores the use of AI to draft consent language from study protocols or IRB applications, examining whether this approach can support regulatory compliance and potentially improve participant comprehension.

Methods: : To examine the potential benefits and risks of using AI in the development of informed consent documents, the project reviewed existing literature, ethical principles, regulatory frameworks, and an ongoing institutional secure Large Language Model (LLM) project locally developed consent AI tool. Sources which include NEJM AI and eClinicalMedicine highlight challenges like literacy barriers and AI's role in potentially improving participants comprehension of complex protocols while ensuring transparency, data security, and scientific accuracy integrity are maintained.

Conclusion: : AI secure platforms such as LLM offer significant potential to enhance the informed consent development that aligns with study protocols, regulation requirements and uses simplified language to improve readability. This initiative provides real-world insight into how AI can be responsibly integrated into consent drafting. Despite AI’s advantages, risks related to confidentiality and scientific validity remain. The effectiveness of AI-assisted consent depends on providing accurate and complete information in the IRB application or protocol. Investigators should also avoid inputting any proprietary or sensitive content from the protocol into the AI platform to avoid compromising data privacy. Using a secure platform and IRB oversight are necessary to ensure data accuracy, non-coerciveness, and ethical integrity of the consent. IRBs should establish guidelines promoting transparency in AI use, require documentation of prompt inputs, and mandate validation of AI-generated consents to protect participant autonomy. AI should complement, not replace, human judgment. This analysis recommends that IRBs cautiously endorse AI-assisted tools for informed consent, while instituting safeguards such as human oversight, transparency, and validations. Structured checklists and proper documentation can support consistent IRB reviews. While AI shows promise, challenges regarding accuracy and contextual appropriateness persist. Empirical studies are needed to evaluate the effectiveness of AI-driven consent forms, develop standardized IRB guidelines and to assess the impact of AI on participant understanding especially among vulnerable populations.

Limitations:: The analysis does not examine all open-source and proprietary AI models and the long-term impact on participant trust. It also does not address possible regulatory changes or errors due to AI-generated hallucinations. Future research should address these gaps and assess the real-world feedback of AI-assisted consent and the impact on IRB review processes.

Discussion: : As AI uses in clinical research is becoming more common, IRBs must ensure their use is transparent, carefully monitored, and safe-guarded by policies. By supporting responsible AI integration, IRBs can encourage innovation while protecting participant rights and ensuring informed consent remains ethically sound. Institutional projects like LLM-based tool offer valuable insights. Future efforts should focus on training reviewers, establishing guidelines, and evaluating AI’s impact on participant understanding. Protecting human participants is the priority and simplifying informed consent language to vulnerable and multilingual populations further upholds the principle of justice, ensuring equal access to understanding and participation in research.