A viral story about ChatGPT supposedly curing a dog’s cancer is the latest case study in how AI achievements get inflated beyond recognition, and why that matters for the entire industry.
The Verge AI reports that the real story behind Australian tech entrepreneur Paul Conyngham’s dog Rosie is far more nuanced than headlines suggest. Rosie wasn’t cured. The mRNA vaccine was given alongside another immunotherapy treatment, making it impossible to know what actually helped. And ChatGPT didn’t design anything, human researchers at the University of New South Wales did the heavy lifting.
What actually happened is still interesting. Conyngham used ChatGPT as a research assistant to parse medical literature and brainstorm treatment options. He worked with university professors to develop a personalized mRNA vaccine for his dog’s tumors. Some tumors shrank. One didn’t respond at all. Conyngham himself acknowledged this isn’t a cure.
Then the hype machine kicked in.
How the Story Mutated
Newsweek ran “Owner With No Medical Background Invents Cure for Dog’s Terminal Cancer.” The New York Post claimed a “tech pro saves his dying dog by using ChatGPT to code a custom cancer vaccine.” OpenAI president Greg Brockman amplified it. Elon Musk jumped in to claim xAI’s Grok also played a role. Social media accounts declared a new era of personalized medicine had arrived.
The gap between reality and coverage is striking:
- What headlines said: ChatGPT cured cancer
- What actually happened: AI helped a non-expert navigate medical literature
- Who did the real work: University researchers with substantial expertise
- The actual result: Partial tumor reduction, not a cure, with unclear attribution
David Ascher, a professor at the University of Queensland, told The Verge AI that AlphaFold “could contribute structural hypotheses about proteins, but it is not a turnkey cancer-vaccine design system.” Alvin Chan, an assistant professor at Nanyang Technological University, said the “AI made this” framing ignores the massive human effort without which “AI’s output would have remained just text on a screen.”
Why This Pattern Keeps Repeating
This story follows a familiar playbook. Someone uses AI tools as part of a larger process involving significant human expertise and resources. The nuance gets stripped away. The narrative becomes “AI did it.” Tech executives amplify the simplified version because it validates their products.
What’s particularly telling is who shared the story and how. Google DeepMind CEO Demis Hassabis shared it without hype. Brockman, whose company made ChatGPT, should have known better but amplified it anyway. Musk used it to promote a competing product.
The incentive structure here is clear, and it’s corrosive. Every inflated claim makes the next round of AI funding easier to justify but erodes public trust in legitimate AI progress.
5 Myths This Story Exposes
- AI can replace medical expertise – It can’t. Rosie’s case required university-level researchers, lab work, and tens of thousands of dollars
- Anyone can do this with a chatbot – Ascher called it “an unusual, highly specific proof of possibility,” not a template ordinary people can reproduce
- The treatment worked – Results are mixed at best, and concurrent treatments make attribution impossible
- mRNA vaccines are proven cancer treatments – They remain largely unproven in humans, let alone dogs
- AI designed the vaccine – AI suggested directions. Humans designed, produced, tested, and delivered the treatment
What AI Practitioners Should Take Away
The real story here is genuinely worth telling. AI tools helped a motivated non-expert engage meaningfully with complex medical literature and connect with the right researchers. That’s valuable. ChatGPT as a research assistant that helps people navigate unfamiliar domains is a legitimate, powerful use case.
But the industry keeps overselling. And every exaggerated headline makes it harder for people to trust AI’s actual, meaningful contributions. For companies building AI products, the temptation to amplify these stories is obvious. The long-term cost to credibility is harder to measure but very real.
As The Verge AI notes, the story “carries a faint whiff of a PR stunt” with “bold claims built from questionable foundations using vague methods” that “comfortably fit inside the world of tech fundraising.”
The AI industry doesn’t need manufactured miracles. The real capabilities are impressive enough, if anyone still believes the messenger.
Full details are available in the original reporting from The Verge AI.