Should You Use AI to Draft Informed Consent?

Darshan Kulkarni and Edye Edens discuss the potential and pitfalls of using AI—like ChatGPT—to draft informed consent documents in clinical research. With both legal and regulatory expertise, they explore how AI could save time, whether it fits institutional IRB requirements, and the real-world value (or lack thereof) for different types of organizations.

Key Takeaways:

• Drafting vs. Final Use: AI can be useful as a first-draft tool, especially for high-volume sponsors. But using it for final documents without oversight is risky.
  
  
• IRB Templates Matter: Many IRBs (especially academic or VA-affiliated) require strict templates—limiting AI's value unless those are integrated upfront.
  
  
• Regulatory Landscape: AI-drafted consents must meet not just FDA standards, but also OHRP requirements. Compliance and clarity are non-negotiable.
  
  
• Customization Is Key: Most current AI tools are just wrappers over ChatGPT. Real ROI comes from domain-specific models trained for clinical research.
  
  
• Data & IP Risks: Feeding protocols into AI raises confidentiality concerns. Plus, who owns the output remains unclear due to copyright issues with training data.
  
  
• Why ROI Falls Short: Companies often reassign internal staff instead of building bespoke solutions. Without a clear use case or strategic planning, results disappoint.

AI shouldn’t replace people—it should support them. Darshan and Edye agree: if you're using AI to generate a first draft of informed consent documents, it could help streamline high-volume workflows. But expecting it to generate a compliant final version is unrealistic. Most current tools feel more like flashy "wrappers" around existing models and often lack a strong ROI. And let’s not forget the hidden risks—like IP concerns and exposing proprietary protocols.

Bottom line? AI has potential, but without strategic investment and oversight, it’s just another overhyped shortcut. Use it wisely, or not at all.

Support the show