If we guard our secrets, patients suffer. Only through a commitment to transparency and data sharing—and artificial intelligence—can we make cell and gene therapies safer and more effective.
Within the complex, living ecosystem of cell and gene therapy, we often find ourselves dazzled by the next big thing — the latest gene-editing tool or a more efficient viral vector. But as someone who has spent decades observing the evolution of this field, I’ve realized that our obsession with individual components is blinding us to a fundamental truth: we are not just building therapies; we are managing an ecosystem.
From my view as founder and CEO of Autolomous, where we build digital manufacturing infrastructure for cell and gene therapy (CGT), I have been struck by how much the industry resembles the early days of aviation. Decades ago, airlines like Pan Am and British Airways were fierce competitors, each guarding their secrets. It took a series of tragedies for the industry to realize a hard fact: if one plane crashes, it doesn’t just affect that airline; it shakes the public’s trust in the entire concept of flying. The solution was the “black box"—a commitment to transparency and shared data that made aviation the safest mode of transport in the world.
Today, CGT is at its “pre-black box” stage, and just like in aviation, safety concerns are thwarting the entire sector. The 2025 deaths from acute liver failure following treatment with Sarepta Therapeutics’ AAV-based gene therapies—along with the interruption in Elevidys shipments and FDA investigation that followed—did not damage Sarepta alone. These events reset the risk conversation for every AAV program in the field, just as two fatal plane crashes, in late 2018 and early 2019, grounded all 387 Boeing 737 MAX aircraft worldwide for 20 months—including those of airlines that had never lost a plane.
We must stop operating in silos and obsessively protecting our intellectual property (IP) while often ignoring the common failures that could teach us how to survive as an industry. We must stop being mere consumers of technology and start being part of a collaborative dialogue.
The key to this transformation lies in the processing power of artificial intelligence (AI), but only if we use it to connect the dots we’ve spent years keeping apart. It’s time to stop looking at AI as a buzzword and start using it as the connective tissue it was meant to be.
The power of “connecting the dots”
Intelligence, in my view, resides in humanity. However, the sheer processing power of modern computers is a prevailing factor that we must leverage to make sense of the vast amounts of data we generate. In CGT, specifically in autologous therapies where treatment is created from the patient for the patient, each individual creates their own “book” consisting of hundreds of pages and thousands of data points.
When we talk about the potential of AI, the most compelling angle isn’t just the math, it’s the connection. I envision a three-legged stool of data capture that must be integrated for our field to thrive. If one leg is imbalanced, the stool falls. The three legs are:
- Scientific data: The information that takes us from inception to clinical and commercial viability.
- Manufacturing data: The real-time capture of what happens during the actual creation of the therapy.
- Post-treatment data: Long-term tracking of efficacy and patient outcomes.
Currently, we treat these as discrete silos. AI’s true potential is to act as the trough between these data lakes, allowing post-treatment results to inform scientific process optimization, which in turn streamlines manufacturing. But this requires data sharing—something that is still sorely lacking in the sector. Fortunately, that is now beginning to change. Regulators are pushing for greater data sharing and parts of the industry are building pre-competitive consortia around it. Still, however, the debate over what gets shared, and on whose terms, is still unresolved.
Why we are falling behind
Why has it taken so long for biotech to catch up with early AI adopters? It is a common question, and the answer is simple: We are still trying to use flat-head screwdrivers on star-shaped screws.
For years, CGT has borrowed tools from traditional pharma. But pharma processes are vastly different. In traditional drug manufacturing, you don’t need to know the viability of your cells every second. In CGT, we do. We need live, momentary data, yet we are still manually capturing information on pen and paper or using 16 different kits to get a single data point.
The early adopters in other industries didn’t just add AI to their existing workflows; they redesigned their workflows for a digital-first world. We, conversely, have been trying to optimize processes that weren’t fit for our purpose in the first place.
Common mistakes for the AI-new biotech
As more companies race to integrate AI, I see the same pitfalls repeated. The most prevalent is the zero-sum game mentality. We are so focused on winning our individual race that we refuse to share the non-IP experiences—our failures. No one goes to the patent office to register a mistake, yet those mistakes are precisely what our peers need to know to avoid wasting years on dead-end research.
Another significant mistake is ignoring the “fourth leg” of data: genetic elements that can be identified even before disease manifests. The field focuses heavily on oncology and post-birth rare disorders, but we have half a century of digital data that could help us correlate family history with disease outcomes. If we ignore this, our AI models are only seeing half the picture.
Finally, there is the over-reliance on off-the-shelf AI without a clear purpose. Using a 24-wheel lorry to transport three people is overkill. You need to identify the right tool for the specific job. I often tell my team, “Don’t just be a consumer of technology.” You must sit at the table with the scientists to define what optimization looks like for our specific biological challenges.
The path forward: A call for collaboration
The ultimate goal of our mission is simple: save lives. To achieve equity of access and make these therapies available worldwide, we must make our processes leaner and slicker. AI is the engine that can get us there. Data is the fuel.
We must learn to collaborate like the aviation industry has. We need an ecosystem where data feeds back through the entire lifecycle of a therapy. AI is coming online at the exact moment we desperately need this connection. It is not just a tool; it is the bridge that will allow us to break down our silos and finally turn the “black box” of CGT into a transparent, thriving reality.
