ChatGPT did not cure a dog’s cancer

When an Australian tech entrepreneur with no background in biology or medicine said ChatGPT helped save his dog from cancer, the story spread with the kind of validation Big Tech has long craved: proof that AI will revolutionize medicine and take on one of its deadliest diseases. The reality, as usual, is more complicated.

ChatGPT did not cure a dog’s cancer

A sick dog, desperate owner, and a bunch of chatbots made for a great story. The actual science was much messier.

The version of the story that made the rounds online, first reported by The Australian, was relatively straightforward. In 2024, Sydney-based Paul Conyngham learned that his dog Rosie had cancer. Chemotherapy slowed the disease but failed to shrink the tumors. After vets said “nothing could be done” for the Staffordshire bull terrier-shar pei, Conyngham said “I took it upon myself to find a cure.”

Conyngham said he used ChatGPT to brainstorm treatment ideas. The chatbot surfaced immunotherapy as an option and pointed him toward experts at the University of New South Wales, who then genetically profiled Rosie’s cancer. He then used ChatGPT and Google’s protein structure AI model AlphaFold to help make sense of the results. With the help of UNSW professor Pall Thordarson, he pursued a personalized mRNA vaccine tailored to Rosie’s tumor mutations. Thordarson told The Australian he thinks it’s the first time such a treatment has been designed for a dog.

A few weeks after Rosie’s first injection last December, Conyngham said her tumors had shrunk and she’s doing better, even chasing rabbits in the park. They’ve not disappeared entirely, though, and one tumor didn’t respond at all. “I’m under no illusion that this is a cure, but I do believe this treatment has bought Rosie significantly more time and quality of life,” Conyngham told The Australian.

That nuance was lost as the story spread. Newsweek ran the headline “Owner With No Medical Background Invents Cure for Dog’s Terminal Cancer,” while the New York Post declared that a “Tech pro saves his dying dog by using ChatGPT to code a custom cancer vaccine.” On social media, many accounts hyped Rosie’s case as a “cure” and a sign a new era of personalized medicine had arrived. Some, notably OpenAI president and cofounder Greg Brockman, should have definitely known better, and others, like Google DeepMind CEO Demis Hassabis, did and shared it without hype. Elon Musk joined in too, keen to point out that xAI’s Grok also played a part — a detail that was absent from much of the original coverage.

The story also gives AI far too much credit. Not only was Rosie not cured of cancer, it’s not clear the mRNA vaccine was responsible for her improvement. The personalized treatment was administered alongside another form of immunotherapy known as a checkpoint inhibitor, designed to help the immune system target tumors, making it difficult to know if the vaccine had any effect at all. One of the scientists involved, Martin Smith, said the team is performing tests to check the immune response.

ChatGPT did not design or create Rosie’s treatment; human researchers did.

Nor was the vaccine itself generated by a chatbot. ChatGPT did not design or create Rosie’s treatment; human researchers did. At most, the chatbot served as a research assistant helping Conyngham parse medical literature — impressive, but a far cry from the breakthrough implied.

Reports are also vague on AlphaFold’s role. David Ascher, a professor and director of biotechnology programs at the University of Queensland in Australia, told The Verge that the model “could contribute structural hypotheses about proteins, but it is not a turnkey cancer-vaccine design system.” Official guidance, he noted, also warns that AlphaFold is not validated for predicting the effects of some mutations and does not model “several biologically important contexts” either.

Grok’s contribution is even harder to pin down. On X, Conyngham wrote that “the final vaccine construct for Rose was designed by Grok,” but it’s not clear what that means in practice or what inputs the model was given. Ascher said Grok would realistically fall into much the same category as ChatGPT: a tool that “could help with literature search, summarising papers, translating jargon, suggesting workflows, drafting code or documents, and helping a user think through options.” A useful role, but hardly what designing a cancer vaccine suggests.

The “AI made this” framing ignores this massive human effort, without which the “AI’s output would have remained just text on a screen.”

All in all, Ascher said Rosie’s case “is better seen as an unusual, highly specific proof of possibility than a template ordinary people can readily reproduce.” It needed “substantial” expert labor, he said, “not just a chatbot and a few prompts.”

That distinction is especially important in medicine, where success depends not just on producing plausible information, but on the expert, physical work of producing, testing, and delivering actual treatment. Alvin Chan, an assistant professor at the Nanyang Technological University in Singapore who is building AI for biomedical and drug discoveries, told The Verge the “AI made this” framing ignores this massive human effort, without which the “AI’s output would have remained just text on a screen.” In Rosie’s case, AI is better understood as a tool for sketching a blueprint than as the creator of the treatment itself.

The whole thing carries a faint whiff of a PR stunt that is hard to shake. Bold claims built from questionable foundations using vague methods comfortably fit inside the world of tech fundraising. mRNA vaccines — much like the broader promise of personalized medicine — remain largely unproven as cancer treatments in humans, let alone dogs, and while the case may be real, it feels too tidy and conveniently glosses over the tens of thousands of dollars and significant expertise required to turn the idea into a viable treatment.

I reached out to Conyngham asking for a chat on X but have not received a response yet. His profile says “Ending Cancer for Dogs” and links to a Google form describing his “dream to make this process something everyone could have access to.” The form asks whether your dog has cancer, whether you’re a researcher or scientist who wants to get involved, and whether you are an investor.

I think it would be a mistake to dismiss Rosie’s story as completely meaningless. AI may not be replacing the lab anytime soon, but it is making science more accessible to ordinary people. However, that’s not the same as making care more accessible, and few patients — or pet owners — have ready access to the world-class experts, specialized equipment, and substantial funds needed to turn that information into real treatment.