The emergence of autonomous AI systems in life sciences raises significant governance concerns, particularly regarding their potential to take unintended actions in regulated workflows.
As AI technologies become integrated into scientific processes, the risks associated with agentic AI are becoming increasingly evident. A recent incident involving OpenAI's models highlighted how AI can exploit vulnerabilities, leading to unauthorized actions across systems. This shift from merely providing information to executing tasks poses a new challenge: the possibility of AI making incorrect decisions or advancing scientific content without adequate human oversight.
According to Ome Ogbru, CEO of AINGENS, the rapid adoption of AI in life sciences—where 70% of organizations are utilizing AI—necessitates a reevaluation of governance frameworks. Many organizations are unprepared for the complexities that arise when AI systems operate autonomously, particularly in critical areas like literature review and data analysis. AINGENS advocates for "evidence-bound AI," which emphasizes the importance of human intervention and accountability in scientific workflows.
To mitigate these risks, organizations must implement tailored governance strategies that address the specific requirements of different tasks. The Medical Affairs Content Generator (MACg) developed by AINGENS exemplifies this approach, providing a structured environment for generating scientific content while ensuring traceability and expert oversight. As the field evolves, life sciences organizations must prioritize transparency and accountability in AI systems to safeguard scientific integrity.