Comment from Dr. Ajay Babu Pazhayattil

Line # 13-15: "This guidance provides recommendations to sponsors and other interested parties on the use of artificial intelligence (AI) to produce information or data intended to support regulatory decision-making regarding safety, effectiveness, or quality for drugs." However, below are two examples where AI solutions can be used for regulator decision-making beyond patient safety, drug effectiveness, or product quality and are still critical in nature: - AI-optimized inventory/supply chain management (which can be submitted to the FDA to prove the availability- and effectiveness of drug shortage/supply security measures) - Generative AI-based document writing (the document can be submitted-eCTD to the FDA for approval review) Recommend re-wording: "This guidance provides recommendations to sponsors and other interested parties on the use of artificial intelligence (AI) to produce information or data intended to support regulatory decision-making." This statement captures other critical regulatory decision-making AI tools beyond- patient safety, drug effectiveness, or product quality AI tools and eliminates the need for yet another guidance document. Line # 249 However, for this example, a manufacturer, as a part of release testing, would measure fill volume on a representative sample for each batch. Measuring fill volume through release testing would reduce the AI model influence, and therefore, the model influence would be determined to be low. Having a representative sample tested does not reduce risk as it is not an absolute control. Observed fill level/calculated fill volume are outcomes-attributes; through applying technology, the current statement reflects the outdated 1987 PV guidance approach of reliance on quality attributes vs establishing input parameter based controls. Fill level variability is, in fact, due to variability of input materials (including ex. vial dimensions), filling equipment calibration, filling process parameters (speed, pressure..), environmental factors prior to sealing, etc. Tight controls over these elements is what ensure minimal fill volume variability, more than the one-time release testing. Recommend re-wording: However, for this example, a manufacturer, as a part of manufacturing controls, has input material controls, continual monitoring and variability control of critical filling process parameters, routine calibration and line qualification. In addition, routine in-process and final release testing are performed. The existing controls, therefore, would reduce the AI model influence, and the model influence would be determined to be low. This statement reflects the current thinking and guidance. Dr. Ajay Babu Pazhayattil