AI in CGT needs AI-worthy data, not just better models

“The datasets underpinning CGT workflows, and the infrastructures that generate and connect them, must become AI-worthy before AI can matter at scale.” AI is widely positioned to transform cell and gene therapy (CGT), but its impact is limited less by an algorithmic deficit than by an AI-worthiness’ deficit: the field’s fragmented and poorly documented datasets are rarely governed, interoperable, and provenance-aware enough to support trustworthy AI. Drawing on the AI Act and the European Health Data Space, this piece argues that CGT developers, manufacturers, and clinical institutions should prioritize disciplined data stewardship now, because the datasets underpinning CGT workflows must become AI-worthy before AI can matter at scale.