AI-assisted dog cancer vaccine goes viral: what actually happened with Paul Conyngham and Rosie
Mar 16, 2026
Key Points
- Australian entrepreneur Paul Conyngham used ChatGPT to design a custom mRNA vaccine for his dog's mast cell cancer, reducing one tumor by half after months of regulatory navigation.
- The viral story exposed a discourse gap: OpenAI framed it as AGI validation while biomedical experts noted mRNA vaccine design is straightforward, and Stripe CEO Patrick Collison clarified ChatGPT functioned as a search tool, not cure-generator.
- AI is compressing the timeline for complex biomedical research from years to weeks, blurring the line between professional institutions and motivated individuals and raising acute questions about FDA approval frameworks for accessible biotechnology.
Summary
An Australian tech entrepreneur named Paul Conyngham used ChatGPT to design a custom mRNA vaccine for his dog Rosie after she was diagnosed with mast cell cancer in 2024. The story went viral over the weekend, triggering a sharp divide between hype amplification and regulatory skepticism.
Conyngham has 17 years of machine learning and data analysis experience and previously directed the Data Science and AI Association of Australia. He took Rosie's tumor DNA and healthy DNA to the University of New South Wales for sequencing, ran the sequences through data pipelines to identify mutations, and used algorithms to find candidate drugs. When a pharmaceutical company refused to supply an existing immunotherapy drug not approved for that indication in dogs, he worked with the university's RNA Institute to design a bespoke mRNA vaccine based on his algorithmic analysis.
After several months navigating regulatory approval, Conyngham administered the vaccine to Rosie. One tumor shrank by half and Rosie's quality of life improved significantly, though she was not completely cured. Cancer cells divide continuously and complete eradication is rarely achieved in a single intervention.
Responses split sharply. OpenAI president Greg Brockman quoted the story as evidence of AGI opportunity. Biomedical engineer Patrick Heiser countered that making single-arm mRNA vaccines is trivially easy. Stripe CEO Patrick Collison clarified that ChatGPT acted as a high-powered search tool, not a cure-generator, a distinction that mattered as the narrative took flight.
Chemist and AI biotech researcher Ash Gilgalacar invoked Freeman Dyson's 2007 essay on domesticated biotechnology. If AI reduces the cognitive overhead required to navigate biological knowledge and assemble complex research pipelines, the boundary between professional research and motivated individuals may blur. That shift carries genuine safety and governance risks but also opens the possibility that biological design becomes something like a creative craft practiced not only by institutions but also by curious individuals experimenting at smaller scales.
The regulatory question is acute. Families facing terminal or deteriorating conditions often have treatment options available but patients do not qualify for trials or access. The FDA has explored right-to-try frameworks in certain scenarios, but the traditional approval process may need recalibration as biotech knowledge becomes more accessible. Conyngham's case demonstrates the efficiency gain clearly. He is a high-agency person with relevant expertise conducting research that would have been possible a decade ago but now takes weeks instead of months. The underlying science is not novel, but the speed is real.