Comment from Anonymous

There are many risks to utilizing AI when it comes to regulatory decision-making, let alone creating regulations for drug and biological products. Previous comments have mentioned the risk of bias in the use of these AI models, and I am inclined to agree. The FDA must address these concerns before implementing AI in their regulatory decisions. The first point of concern is data integrity and quality. AI models utilize existing data online, which can lead to biased outcomes. There also may be a lack of transparency in how AI models process data, which could undermine confidence in results. Also, the draft guidance proposes a risk-based credibility assessment framework for a particular context of use (COU). What exactly does this framework entail, other than defining credibility evidence? How will this framework be applied? Considering that AI models will be used even in the clinical phases of the drug product life cycle, it is important to allow transparency to ensure that such methods were used. The development of additional guidance regarding pharmacovigilance would prove beneficial to allow transparency of the use of AI models in regulatory decision-making. These can include the clear documentation of AI model architecture and training data sources, conducting model performance assessments, and applying fairness metrics to avoid overlooking safety concerns. Implementing these measures can effectively allow the FDA to further support regulatory decision-making for industries.