Explaining The Unexplainable: How To Win US FDA’s Trust In Artificially Intelligent Drug Manufacturing

One issue that has surfaced as the pharmaceutical industry responds to the US FDA’s recent discussion paper on manufacturing with AI machine learning models is the question of how to explain what the models are doing.

Artificial and human intelligence
the decisions are obvious even when the explanations are not • Source: Shutterstock

The question of how to explain what artificial intelligence models are doing is emerging as a key regulatory issue in pharmaceutical manufacturing.

The “explainability” of self-learning AI models can be so low, many would sidestep the challenge by treating

Rather than try to explain the models, they would constrain them with careful training and guardrails – and by

More from Innovation

More from In Vivo

Digital Transformation At The FDA: Generative AI Set To Transform Drug Review Process

 
• By 

The FDA plans to implement generative AI for drug reviews by 30 June 2025, enhancing efficiency and potentially accelerating approval processes. Discussions with OpenAI about AI integration are ongoing.

Deals Shaping The Industry, April 2025

 
• By 

An interactive look at pharma, medtech and diagnostics deals made during April 2025. Data courtesy of Biomedtracker.

Plans For London Cancer Hub Promises New Destination For Biotech Innovation

 
• By 

Plans have been submitted for a £1bn development in London that can home biotech companies and encourages cross collaboration with the Institute of Cancer Research. In Vivo took an exclusive tour of the UK’s prospective new cancer innovation district.