Engineers use AI to design an mRNA vaccine to treat rescue dogs' cancer; the tumor shrank by half but was not cured, and regulatory approval has become the biggest obstacle.

Gate News reports that on March 16, Sydney-based machine learning engineer Paul Conyngham used ChatGPT and DeepMind’s AlphaFold to self-educate in mRNA vaccine design, creating a custom vaccine for Rosie, a rescue dog with mast cell cancer. He collaborated with the University of New South Wales (UNSW) RNA Research Institute to complete the vaccine design and administered the injection at the University of Queensland Gatton Veterinary School. After the injection, Rosie’s tumor significantly shrank, and veterinarian Professor Paola Allavena said, “The tumor probably shrank by about half.” This story has recently been widely shared on social media as “AI cured the dog’s cancer.”

However, according to the original report from the University of New South Wales, Rosie’s cancer is still progressing, and a cure is still a long way off. Biomedical engineer Patrick Heizer pointed out that manufacturing a single mRNA vaccine is “extremely simple” technically, but the real challenge and expense lie in proving the vaccine’s safety and efficacy through randomized controlled trials, which has not yet been accomplished.

Conyngham himself stated that the biggest obstacle throughout the process was not vaccine design but ethical approval: he spent three months, working two hours every night, to write a 100-page ethics approval document, “more difficult than making the vaccine.” Biological author Ruxandra Teslo, citing a similar experience of GitLab co-founder Sid Sijbrandij (who self-funded experimental therapy after osteosarcoma relapse and has not relapsed since 2025), believes that regulatory bureaucracy in early clinical trials is the core bottleneck hindering the implementation of personalized medicine.