An Australian IT professional created a cancer vaccine for his dog using ChatGPT

Paul Coningham, a Sydney-based entrepreneur with 17 years of experience in machine learning, learned that his dog Rosie had months to live. Rather than accept the diagnosis, he did what he does best: build a pipeline from data. The result is the world’s first personalized mRNA cancer vaccine designed for a specific dog. And it worked.

«When she was sentenced, I felt I had to do everything I could,» Coningham told Australia’s Today program.

How it all began
In 2019, Coningham adopted Rosie, a Staffordshire Bull Terrier and Shar-Pei mix, from a shelter. A few years later, the dog was diagnosed with aggressive mast cell cancer. Traditional treatments — surgery, chemotherapy, immunotherapy — slowed the spread of the disease, but the tumors didn’t shrink.

A pharmaceutical company refused to provide a promising drug. Coningham did not stop.

Pipeline instead of recipe

The first step was to sequence the tumor’s DNA through the Ramaciotti Centre for Genomics at the University of New South Wales (UNSW) — a process that cost around A$3,000. By comparing Rosie’s healthy DNA with the tumor’s, he identified harmful mutations.

Next — ChatGPT as a research assistant: helped navigate the biomedical literature, plan the analysis steps and pointed to the necessary specialists at UNSW. AlphaFold from Google DeepMind was used to model 3D structures of mutated proteins and identify potential therapeutic targets. Own ML algorithms helped select neoantigens — mutated proteins that are most likely to provoke an immune response in the body.

The result was a half-page mRNA sequencing protocol. UNSW RNA Institute Director Pall Thordarson produced a physical vaccine in less than two months.

Regulatory approvals took three months, with Coningham spending two hours every night filling out a 100-page document. The bureaucracy, he says, proved more difficult than the development itself.

Rosie received her first injection in December 2025, with subsequent boosters. By mid-March, one of the tumors had shrunk by about 75%.

In January 2026, Rosie jumped over a fence to chase a rabbit—a behavior that had disappeared during her illness, returning. Coningham considers this the most significant data point of the entire project.

What scientists say

The reaction from the scientific community has been one of cautious delight. UNSW Associate Professor Martin Smith couldn’t hide his excitement: «It was like, holy shit, it worked!» UNSW structural biologist Kate Mickie said it was «encouraging to see someone without a degree be able to do a pipeline like this.»

OpenAI President Greg Brockman publicly called this case study «the first personalized cancer vaccine designed for a dog.»

Scientists also note that similar techniques could potentially form the basis of personalized therapies for humans.

But there is an important caveat
Coningham himself stresses: «I have no illusions that this is a cure.» This is a single result, not clinical evidence — one tumor in one dog responded to one vaccine. And this was not «a casual user with a chatbot,» but a technically savvy human with an ML background, supported at every critical step by skilled scientists from UNSW.