Image: Stephen Wang Photography / Torigen Pharmaceuticals
In mid-March 2026, Paul Conyngham’s story went viral: a Sydney-based AI consultant with no medical training reported using ChatGPT, Grok, Gemini and AlphaFold to help design a personalized mRNA cancer vaccine for his dying dog. The coverage spread fast with at times hyperbolic claim: CBS, Fortune, Newsweek and dozens more. OpenAI CEO Sam Altman called it “an amazing story” to his millions of followers. DeepMind CEO Demis Hassabis shared it too.
Rosie, an eight-year-old Staffordshire Bull Terrier–Shar Pei mix, had reportedly been misdiagnosed for nearly a year before a biopsy confirmed aggressive mast cell cancer in 2024. Surgery and chemotherapy did not stop the disease. “She was beyond treatment, so I had to do whatever I could,” Conyngham told CBS News. “And whatever I could was AI. That’s what I know.”
The hook was irresistible: man saves dog with ChatGPT. But the real story is more complicated. “It’s super cool, but I think that it’s going to be a challenge to really understand how that could be applied to your everyday dog patient,” said Ashley Kalinauskas, CEO of Torigen, a veterinary cancer immunotherapy company that has treated roughly 2,000 animals since spinning out of Notre Dame in 2017.
Even a simplified version of what Conyngham did, Kalinauskas said, would take months of focused work. And critical clinical details remain unclear from the outside: the specifics of Rosie’s diagnosis, including her tumor grade, were not detailed in much of the media coverage. Grade matters significantly for outcomes. For grade two and below, surgery alone is curative about half the time.
The underlying mRNA platform uses the same broad modality behind the Pfizer-BioNTech and Moderna COVID-19 vaccines, products that underwent trials involving tens of thousands of participants and still faced sustained public skepticism about their safety and efficacy.
Who else is in the field
Ashley Kalinauskas
Veterinary oncology does not have the funding infrastructure that drives human drug development. Clinical studies are smaller, regulatory pathways vary by product class, and much of the evidence base is still early-stage.
Torigen’s autologous whole-cell approach is one of several emerging options. ELIAS Animal Health holds USDA approval for a personalized vaccine-plus-T-cell therapy for canine osteosarcoma. Vaxkit distributes a European clinic-prepared autologous immunotherapy in North America. Boehringer Ingelheim’s Oncept targets canine melanoma with a DNA vaccine. Each takes a different regulatory route and a different immunological approach.
A shifting regulatory framework
The USDA’s framework for autologous veterinary biologics has shifted significantly in recent years, and companies in the space have had to shift with it. When Torigen first brought its product through the process, it fell under 9 CFR 103.3, an experimental authorization pathway. Kalinauskas said the arrangement required Torigen to reapply every two years with new efficacy data to stay on the market.
“A lot of things they didn’t know initially, how to regulate these products,” Kalinauskas said. “And there were a lot of questions as to how we want to regulate them, and what safety and efficacy testing are going to be needed in order to get these products approved.”
The biennial reauthorization created uncertainty. “From a startup company, it actually was a pretty big risk,” she said, “because essentially, what are they looking for every two years in order to renew my authorization?” For investors doing due diligence, the structure was a red flag.
Torigen eventually moved to a different regulatory classification, 9 CFR 107.1, which covers in-clinic preparation kits. Under that framework, the company sells the kit; the veterinarian uses it to prepare the product on-site for their own patient. “By selling a kit, it allows for the veterinarian to do the in-clinic preparation,” Kalinauskas said. “So it’s just under a different umbrella.”
Under 107.1, VIM Clara is exempt from standard product and establishment licensing when prepared by a veterinarian. It is classified as a veterinary-use in-clinic preparation kit, not a licensed drug or vaccine. The vet makes the product; the exemption applies.
What exists today
Not long ago, Kalinauskas said, the standard of care for most pet cancers was “surgery, and then prednisone and prayers.” Pet owners today want more than that. “Most owners are saying, I don’t want to put my pet on chemo,” she said. The market is responding, but the evidence, across the board, is still being built.
Torigen’s product, VIM Clara, is an in-clinic kit that lets veterinarians create a personalized immunotherapy from a patient’s own surgically excised tumor tissue in roughly 40 minutes, for dogs, cats and horses. The company has raised more than $16 million. Its approach preserves the whole tumor cell surface rather than targeting specific mutations. “We want to present that whole fingerprint in order to have the body say, this is what we need to drive a response against,” Kalinauskas said. “An mRNA-based vaccine is going to be a small snippet of that.”
Torigen’s published clinical data is early-stage but growing. In a 2020 BMC Veterinary Research study, the vaccine arm included eight dogs with metastatic hemangiosarcoma; median survival was 142 days versus 41 days for surgery alone. Kalinauskas said Torigen also has mast cell tumor data on 14 patients and an unpublished equine study of 15 patients showing a 40% objective response rate and 73% clinical benefit. Safety data covers over 100 dogs, roughly 100 cats, and about 40 horses. Long-term outcome tracking has been difficult: the clinic follow-up response rate (the rate at which treating veterinarians report back on patient outcomes) has been under 2%.
“There is not going to be one tool that hits everything,” Kalinauskas said. “It is about the best combination to prevent the tumor from coming back.” The ideal approach, she acknowledged, would combine both autologous whole-cell and targeted mRNA methods. But from a regulatory perspective, she said, “that will not be possible in the United States.” The cost to pet owners for Torigen’s therapy is typically under $2,000. Kalinauskas has also written that the ideal protocol would add a checkpoint inhibitor, the same class of drug Rosie was receiving alongside her mRNA vaccine.
An emerging field
The science of autologous immunotherapy is grounded in real mechanisms: presenting tumor antigens to the immune system to provoke a targeted response. But the clinical evidence that it works reliably, at scale, in veterinary patients is still accumulating. That creates a gap between how these products are described and what has been proven, and the gap runs through the entire veterinary immunotherapy space.
Torigen’s website tells pet owners its therapy “teaches the immune system to attack cancer cells” and can “train the immune system to find cancer cells in your pet’s body and kill them.” That language describes what the product is designed to do. But the same site, in its fine print, says: “Safety and efficacy have not been established.”
Kalinauskas was candid about the limits of the evidence in conversation, more precise than the marketing copy suggests. That’s not unusual for a company in an emerging therapeutic category, but it is something a pet owner reading the website deserves to know.
The uptick in demand
The Rosie story generated more than media coverage. It generated demand, including for Torigen. “People reach out and say, wait, you do personalized cancer vaccines. What is the difference?” Kalinauskas said. “So we put a blog post on our website explaining the difference between an mRNA-based vaccine and the autologous vaccine we make.”
Soon after the story went viral, Conyngham launched RosieVaccine.com and opened a waitlist. The site promises to make the breakthrough “accessible” and describes “a clear, guided three-step process” ending in a synthesis-ready mRNA blueprint. At the bottom, in smaller type: RosieVaccine is “an experimental AI-assisted bioinformatics design service.” It does not manufacture vaccines, administer treatments, or practice veterinary medicine. The blueprints it generates are “computational outputs.” And Rosie’s case, the site acknowledges, “represents a single case and should not be interpreted as a promise of similar results.”
“They are still making guesses,” Kalinauskas said of the Rosie approach. “Maybe good guesses, especially if they are hitting driver mutations, but it is still guesswork.”
