Our quest to accelerate vaccine and anti-infective technology development has acquired a new urgency as the biological capabilities of frontier AI models have rapidly advanced, and we enter a new era of engineered pathogens and pandemics.
Until very recently, it took substantial education and practical experience in biology to be able to develop and release a superbug, a background that only a few hundred thousand individuals across the world likely have. Today, individuals without any fundamental understanding of virology or molecular biology can use frontier AI models, ubiquitous tools like ChatGPT, Claude, and Gemini, to substantially close the gap and potentially put themselves on a path to creating a deadly global pandemic for which there are no existing cures. This means we are not only fighting on the already-rapid timescales of biological evolution, but also at the far-accelerated pace of human ingenuity augmented by a generative intelligence trained on the entire corpus of global scientific knowledge, and we must accelerate the development of our defenses now, before it is too late.
Today's pharmaceutical companies are not nimble and agile enough to combat this fast-moving threat. An effective defense requires a full-stack approach: few-shot machine learning-led design coupled with a rapid validation and production process in an adaptable, always-on manufacturing facility.
To do this right, we must build a library of vaccines and other anti-infectives that are already able to combat all relevant families of viral and bacterial pathogens from which potential bioterrorists and the generative models they use draw their inspiration, which we must constantly update based on the expanding biological design capabilities of frontier AI models as they become available for public use.
It is essential that frontier AI labs and governments work together with vaccine and anti-infective developers to build active full-stack defenses against engineered pathogens, but they have only just begun to grapple with this problem, with most funding still going to model capabilities evaluation and red teaming. Without a ‘hot’ pipeline to turn the insights from these exercises into actual medicines, the AI labs expose themselves to massive liability, and humanity runs the very real risk of being left defenceless in the face of even a single malign actor with access to an LLM and a credit card.
PopVax has already taken the first steps towards building this pipeline. With our high-throughput design and immune data generation capabilities, as well as our fast-turnaround GMP-capable RNA Foundry in Hyderabad, we are advancing the frontier of rapid response against de novo pathogens by building first-in-class machine learning models for immunology-aligned vaccine and anti-infective generation tightly integrated with lab-in-the-loop capabilities for testing and validation that extend seamlessly into rapid and flexible manufacturing for these programmable medicines.
If you'd like to collaborate with us on this, email biosecurity [at] popvax.com