FDA official offers tips on leveraging AI in drug manufacturing

Tina Kiang, director of the Division of Regulations and Guidance in the US Food and Drug Administration’s (FDA) Office of Pharmaceutical Quality, emphasized that although artificial intelligence (AI) is a useful tool in many areas of pharmaceutical manufacturing, humans are ultimately responsible for making decisions based on its outputs.

In her presentation at the recent CASSS Well-Characterized Biotechnology Products (WCBP) conference in Washington, DC, Kiang discussed how AI can be integrated into the manufacturing process and its application throughout the entire product lifecycle.

Kiang emphasized that AI can be applied throughout the entire drug product lifecycle, though it must be used in accordance with applicable regulations. AI has applications in drug discovery, nonclinical development, clinical development, manufacturing, regulatory submissions, and in postmarket surveillance.

To help manufacturers incorporate AI in drug development and manufacturing, Kiang mentioned several documents FDA has recently published related to AI, including a discussion paper, scientific article, and draft guidance on the application of these AI tools in pharmaceutical manufacturing.

The discussion paper, released in 2023, was aimed to gather input on several key topics: how cloud applications could impact the oversight of pharmaceutical manufacturing data and records, how the increased volume of data generated by using AI might influence current data management practices, and how the regulatory oversight of AI applications in pharmaceutical manufacturing could be altered or challenged. (RELATED: FDA seeks feedback on artificial intelligence in drug manufacturing, Regulatory Focus 1 March 2023)

This was followed by a 2024 article, An Examination of Process Models and Model Risk Framework for Pharmaceutical Manufacturing, which was published in the International Journal of Pharmaceutics. This paper was authored by FDA staff and officials from the European Medicines Agency (EMA).

The paper examined the use of AI process models as tools for designing and controlling pharmaceutical manufacturing processes. It also investigated a risk-based framework for validating these models and ensuring their maintenance throughout their lifecycle. Furthermore, the paper presented hypothetical case studies to illustrate the implications of applying a model risk framework during the validation of process models.

In January 2025, FDA released draft guidance that addresses the use of AI to support regulatory decision-making. The guidance outlined a risk-based framework for assessing the credibility of AI models and discussed how to establish and evaluate these models for specific contexts of use (COU). (RELATED: AI in drug development: FDA draft guidance addresses product lifecycle, risk, Regulatory Focus 7 January 2025)

From the guidance and papers, several key lessons emerge. She stressed that “AI is a tool and an AI model provides the output. It does not make decisions for anybody. You have to think about AI as any other type of software; it’s very intelligent software, but it is still software. We have to remember that ultimately, decisions are made by people.”

She also added that “the implementation of artificial intelligence must be in accordance with cGMP, that means section 501(a)(2)(B) of the [Federal] Food, Drug, and Cosmetic Act and sections 210 and 211 of the Code of Federal Regulations.”

In addition, she said that AI models should be appropriately validated for the context of use in the manufacturing process and records kept available for inspectors.

She also discussed how manufacturers should assess the risk and level of validation needed to incorporate AI into a process. “When you’re looking at model risk, once you integrate AI into your systems, you have to look at the overall model risk, the decision consequence, and the model influence to determine what level of credibility assessment activities will need to be done before integrating that model into your manufacturing processes,” she said.

Kiang stressed that the quality control unit is ultimately responsible for ensuring the overall quality of the final drug product (21 CFR 210.3), and the quality control unit responsibilities are described in 21 CFR 211.22 and 211.68 for finished drug product.

“People are ultimately the ones who make decisions. You can integrate AI into a process, but ultimately how that AI is used and how the output is used from AI are controlled by the quality control unit during manufacturing,” Kiang said.

FDA official offers tips on leveraging AI in drug manufacturing

Related topics

Recommended content

Welcome to the new RAPS Digital Experience