ChatGPT did not cure a dog's cancer

Australian tech entrepreneur Paul Conyngham had no background in biology or medicine when his dog Rosie was diagnosed with mast cell cancer in 2024. After chemotherapy failed to shrink the tumors and veterinarians said nothing more could be done, Conyngham decided to take matters into his own hands.

What happened next became a viral story of AI-powered medical breakthrough. The reality, according to scientists involved in the case, tells a different tale about the limitations of current AI in healthcare and the dangers of overhyping unproven claims.

The treatment process required tens of thousands of dollars, world-class expertise, and specialized equipment. Conyngham worked with researchers at two Australian universities to sequence Rosie's tumor DNA, identify mutated proteins, and design a targeted mRNA vaccine using AI tools including ChatGPT and Google's AlphaFold protein structure model.

After Rosie's first injection in December 2024, Conyngham reported that her tumors had shrunk and her energy improved. But the story that spread online glossed over crucial details: Rosie wasn't cured, and it's unclear whether the experimental vaccine was responsible for her improvement.

The Science Gets Lost in Translation

As news outlets picked up the story, nuance disappeared. 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." OpenAI president Greg Brockman and Google DeepMind CEO Demis Hassabis shared the story on social media, with Elon Musk adding that xAI's Grok chatbot also played a role.

The viral narrative gives AI far too much credit, according to experts who reviewed the case. ChatGPT did not design or create Rosie's treatment—human researchers did. The chatbot served as a research assistant, helping Conyngham navigate medical literature and brainstorm approaches.

David Ascher, a professor and director of biotechnology programs at the University of Queensland, told The Verge that AlphaFold "could contribute structural hypotheses about proteins, but it is not a turnkey cancer-vaccine design system." The model's official guidance warns it's not validated for predicting mutation effects and doesn't model "several biologically important contexts."

Grok's contribution is even murkier. While Conyngham claimed on X that "the final vaccine construct for Rose was designed by Grok," it's unclear what that means in practice. Ascher suggests Grok likely fell into the same category as ChatGPT: helping with literature searches, summarizing papers, and suggesting workflows—useful, but hardly designing a cancer vaccine.

The Treatment Reality Check

The personalized mRNA vaccine wasn't administered in isolation. Rosie also received a checkpoint inhibitor, a proven immunotherapy designed to help the immune system target tumors. This makes it impossible to determine whether the experimental vaccine contributed to her improvement or whether the established treatment was responsible.

Martin Smith, one of the scientists involved, said the team is still performing tests to check Rosie's immune response to the vaccine. "I'm under no illusion that this is a cure," Conyngham told The Australian, "but I do believe this treatment has bought Rosie significantly more time and quality of life."

Alvin Chan, an assistant professor at Nanyang Technological University who builds AI for biomedical discoveries, argues that the "AI made this" framing ignores the massive human effort required. Without expert researchers, he told The Verge, "AI's output would have remained just text on a screen."

The Bigger Picture Problems

Ascher views Rosie's case as "an unusual, highly specific proof of possibility than a template ordinary people can readily reproduce." The treatment needed "substantial" expert labor, "not just a chatbot and a few prompts."

This distinction matters especially in medicine, where success depends not just on producing plausible information, but on the expert, physical work of producing, testing, and delivering actual treatment. AI may be making scientific information more accessible, but that's not the same as making care more accessible.

The case carries what some observers call "a faint whiff of a PR stunt." Bold claims built from questionable foundations using vague methods fit comfortably within the world of tech fundraising. mRNA vaccines remain largely unproven as cancer treatments in humans, let alone dogs, and the story conveniently glosses over the expertise and resources required.

Conyngham's X profile now 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 who wants to get involved, and whether you are an investor.

Few pet owners have ready access to world-class experts, specialized equipment, and the substantial funds needed to turn AI-generated information into real treatment. While AI tools may be democratizing access to scientific literature, the gap between reading about treatments and actually delivering them remains vast.

The story of Rosie and ChatGPT isn't meaningless—it demonstrates AI's potential as a research assistant and the possibility of accelerating certain scientific workflows. But it's not the medical breakthrough that viral headlines suggested. When tech companies compete for validation through breathless coverage, that distinction matters more than ever.