BIONEWS
If airlines share near-misses, why won't cell therapy share failed batches?
Cell and gene therapies treat tiny patient populations, often for rare cancers, so each failed batch can mean a life rather than a line in a report.
Cell and gene therapies treat tiny patient populations, often for rare cancers, so each failed batch can mean a life rather than a line in a report.
When a cell therapy batch fails quality control, the failure usually stays where it happened: a paper binder in a GMP suite, a spreadsheet on a quality engineer's laptop, a release report filed under a manufacturer's name and never shared outside it. Each individual batch treats a small number of patients, often for rare cancers or single-gene disorders, so the cumulative loss of failure data is a patient-safety problem dressed up as an IT problem.
The industry knows this. In a recent interview with Genetic Engineering & Biotechnology News, Alexander Seyf, CEO of Autolomous, a vendor of digital manufacturing software for cell and gene therapies, called paper-based data management the "elephant in the room" blocking artificial-intelligence adoption. "Everybody wants to have AI. But where do you have your data? If it's in binders, there's not much you can do," Seyf said. On the question of when to start, he was direct: "The sooner you start, the better it is. Pen and paper do not prevail, and pen and paper do not transfer."
The vendor framing is familiar, but the underlying structural problem is bigger than any one supplier's product. A 2026 peer-reviewed review in MDPI Pharmaceutics, "AI and the Transformation of Cell and Gene Therapy Development," is one of several recent independent takes framing digital infrastructure as a precondition for AI in the field (MDPI Pharmaceutics 18(3):356). The UK BioIndustry Association's resource report on cell and gene therapy reaches a similar conclusion from the industry-association side (UK BIA — Cell & Gene Therapy: Driving Efficiencies and Innovation). BioSpectrum Asia's industry coverage has framed digital manufacturing as the necessary lead for the next phase of cell and gene therapy scale-up (BioSpectrum Asia).
So the diagnosis is broadly shared. The harder question is why the obvious fix has not happened, and the answer has very little to do with software.
The first force is incentive. A failed batch is, at the same time, a quality event, a commercial event, and a regulatory event. Documenting it in detail and sharing it with peer manufacturers exposes the disclosing company to competitive, reputational, and historically liability risk, while the benefits of disclosure flow to the field as a whole rather than to the firm that paid the cost. That is the textbook structure for under-investment in collective goods, and cell and gene therapy is a textbook case: batch sizes are tiny, regulatory pathways are tight, and intellectual property around process know-how is a core asset. Seyf's warning that inaccessible or fragmented data causes researchers to lose "downstream scientific opportunities" is the same observation from a different angle. Every batch that dies in a binder is a learning event the rest of the field never gets to have.
The second force is the absence of a neutral sharing infrastructure that would make disclosure safe. This is where the aviation analogy stops being a flourish and starts being a working operational model. In commercial aviation, near-miss and incident reporting systems eventually gave pilots and operators a way to disclose failures without those disclosures becoming enforcement or competitive weapons. Three features made that work: legal protection from regulatory use of the disclosed data, a neutral administrator rather than a competitor receiving the report, and a feedback loop that turned aggregated lessons into operational guidance. Cell and gene therapy manufacturing has no equivalent institution, no equivalent legal protection, and no equivalent feedback loop. The MDPI review and the UK BIA report both situate the field's data culture closer to a closed-shop model than to an open reporting model.
The third force is the publication and IP regime that already shapes what gets out the door. Academic publication is biased toward positive results, and the process patents and trade-secret protections that surround cell therapy manufacturing give firms a defensible reason to keep failure data in-house. Neither of these is going away, and neither has to. What is missing is the kind of structured, protected channel that aviation built, in which a manufacturer can disclose a process failure, an out-of-specification event, or a near-miss without that disclosure becoming a competitive weapon in someone else's hands.
The vendor response to all of this has been to build the digital infrastructure anyway, on a commercial basis. Autolomous has signed a partnership with Center for Breakthrough Medicines to provide digital manufacturing solutions for cell and gene therapy production (Cell & Gene), launched the Digital Pioneer Program with Adva Biotechnology targeting autologous cell therapy manufacturing (Medpath), and announced its own launchpad ecosystem program (Autolomous). Those are real commercial actions, and they are also the obvious objection to taking any one vendor's "digitize or fall behind" framing at face value. A vendor that profits from digitization has a structural reason to call the gap existential, and the legitimate criticism that the industry is failing to digitize does not require accepting that framing as neutral. The MDPI review, the UK BIA report, and BioSpectrum Asia's coverage each reach the digitization conclusion through independent paths, which is what makes the underlying point defensible without leaning on a single supplier.
The question worth watching is not whether cell and gene therapy manufacturers will eventually replace their paper batch records, which they will, but whether the field builds the kind of protected, neutral, multi-party infrastructure that would make failure data shareable at all. The aviation model, in which a neutral administrator handles protected disclosure and feeds lessons back to operators, is one path. A regulatorily mandated disclosure regime is another. A consortium-governed channel run through an industry body such as the BIA is a third. The binders are a symptom. The work is the institution that has to replace them.