In 2024, Sydney tech entrepreneur Paul Conyngham confronted a grim veterinary prognosis: his dog Rosie had cancer that persisted despite chemotherapy and surgery. He turned to artificial intelligence and academic partners, commissioning genomic sequencing and, with university scientists, building a bespoke mRNA cancer vaccine. The effort has coincided with visible clinical improvement for Rosie, even as some tumors remain.
A personalized mRNA vaccine built at speed
Conyngham, an electrical and computing engineer who cofounded Core Intelligence Technologies and previously served as a director at the Data Science and AI Association of Australia, sought out the University of New South Wales for tumor sequencing and analysis. OpenAI’s ChatGPT suggested immunotherapy and pointed him to campus capabilities, while AlphaFold-an AI system from DeepMind-was used to examine mutated proteins for potential immune targets. Pall Thordarson, director of UNSW’s RNA Institute, led the rapid design of a custom mRNA construct intended to prime Rosie’s immune system against tumor-specific antigens.
“This is the first time a personalized cancer vaccine has been designed for a dog,” Thordarson said. “This is still at the frontier of where cancer immunotherapeutics are-and ultimately, we’re going to use this for helping humans. What Rosie is teaching us is that personalized medicine can be very effective, and done in a time-sensitive manner, with mRNA technology.” The case also underlines how consumer-facing AI tools can now act as informal front doors into complex research and regulatory systems that were previously navigated almost exclusively by clinicians and scientists.
What happened and when
- 2024: Rosie diagnosed with cancer; chemotherapy and surgery undertaken but tumors persisted.
- Mid-late 2024: Genomic sequencing performed; AI tools used to prioritize potential neoantigens; a conventional immunotherapy option was identified but not supplied by the manufacturer.
- December 2025: First injection of a bespoke mRNA cancer vaccine administered under veterinary supervision.
- February 2026: Booster dose delivered, with clinicians and researchers continuing to monitor safety signals and tumor burden.
Early outcomes in a single animal
- Tumor response: Most tumors have shrunk substantially; some have not responded, highlighting the heterogeneity typical of advanced cancers.
- Clinical status: Energy and activity have improved. “In December she had low energy because the tumors were creating a huge burden for her,” Conyngham said. “Six weeks post-treatment, I was at the dog park when she spotted a rabbit and jumped the fence to chase it. 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.”
- Durability: Duration of response and long‑term control are unknown and will depend on continued monitoring, follow‑up imaging, and repeated clinical assessments.
- Limitations: This is n=1, outside a controlled trial, with prior chemotherapy and surgery that may confound attribution of benefit. No regulatory body has evaluated this bespoke construct for safety or efficacy, so the outcome remains an anecdotal signal rather than clinical evidence.
How an mRNA cancer vaccine is tailored
- Identify mutations: Tumor DNA/RNA are sequenced to find alterations unique to the cancer, distinguishing malignant cells from normal tissue.
- Prioritize neoantigens: Computational tools score which mutations are most likely to be seen by the immune system and to generate a strong T‑cell response.
- Encode into mRNA: Selected sequences are encoded into an mRNA template that instructs cells to produce the targeted antigens, effectively turning the patient’s own cells into short‑term vaccine factories.
- Deliver safely: The mRNA is packaged-commonly in lipid nanoparticles-to reach immune cells and prompt a T‑cell response, with dose and route of administration adjusted to the species and tumor type.
- Iterate: Boosters or updated constructs may be used if the tumor evolves or initial targets are insufficient, creating a potential cycle of “adaptive” vaccine redesign.
AI’s role-and its boundaries
- Clinical concepting: “I went to ChatGPT and came up with a plan on how to do this,” Conyngham said, underscoring how generative AI can rapidly synthesize public-domain oncology concepts for lay users.
- Target discovery: Protein-structure models such as AlphaFold can help prioritize mutated proteins for vaccine design, but laboratory validation remains essential before any construct is manufactured and administered.
- Safeguards: AI systems are not medical devices; any output requires expert oversight, ethics review, and quality-controlled manufacturing before use in animals or humans. As regulators consider how to govern AI in clinical workflows, cases like Rosie’s will test where advisory use ends and de facto medical decision‑making begins.
Where this sits in the wider oncology landscape
- Evidence base: Personalized neoantigen vaccines are advancing in human oncology trials, particularly in melanoma and other solid tumors. Early data suggest they can amplify anti-tumor T‑cell responses, often combined with checkpoint inhibitors, but large randomized trials and longer follow‑up are still needed.
- Comparative oncology: Dogs develop spontaneous cancers that share biology with human tumors, making them important translational models for immunotherapy research and a potential bridge between lab studies and human phase 1 trials.
- Access and capacity: Rapid vaccine design depends on local genomics, bioinformatics, and RNA manufacturing capability-resources concentrated in a limited number of institutions. Policymakers weighing investments in “sovereign” mRNA and sequencing capacity will increasingly need to decide whether and how veterinary applications are included in those national platforms.
Regulatory and ethical guardrails
| Domain | Typical oversight/practice | Relevance to this case |
|---|---|---|
| Veterinary biologics (Australia) | Regulated as veterinary medicines; institutional Animal Ethics Committee (AEC) approval required for research use in animals, alongside national codes for animal care and use. | University involvement implies AEC review for animal welfare; bespoke products fall outside routine commercial registration, raising questions for regulators about how to treat one-off, AI‑assisted constructs. |
| Veterinary biologics (United States) | USDA’s Center for Veterinary Biologics licenses and monitors veterinary vaccines and autogenous biologics, setting standards for potency, safety, and purity. | Provides a reference model for oversight of custom animal vaccines in large markets, including how far to permit personalization before a product must enter formal review. |
| Human therapeutic vaccines (Australia) | The Therapeutic Goods Administration regulates clinical trials, manufacturing quality (GMP), and market authorization for human vaccines and biologics. | Any translation to people would require phased clinical trials with strict safety and efficacy standards, as well as pharmacovigilance and post‑market surveillance frameworks. |
| Human therapeutic vaccines (United States) | FDA’s Center for Biologics Evaluation and Research oversees trials and licensure of mRNA vaccines and immunotherapies under existing biologics regulations. | Highlights the gap between promising animal experiences and human approval pathways, where AI‑enabled personalization will have to fit inside long‑standing rules on risk, consent, and manufacturing quality. |
System-level implications to watch
- Timelines: The vaccine was designed and produced in under two months, underscoring how genomic workflows and mRNA platforms can compress R&D cycles. For health systems and regulators, this raises the prospect of “on‑demand” biologics that move faster than traditional review processes.
- Manufacturing quality: Scaling bespoke vaccines will hinge on GMP‑grade RNA production, sterility assurance, and batch documentation. Without clear standards for micro‑batch production, regulators will face pressure to adapt rules designed for large commercial runs.
- Data standards: To inform policy, cases like Rosie’s need systematic tumor measurements, immune profiling, and time‑stamped outcomes shared under research protocols, rather than isolated social‑media narratives.
- Equity: Personalized approaches remain expensive; without dedicated funding models, access will be narrow and uneven. Governments and insurers will need to decide whether such interventions are treated as experimental research, elite boutique care, or candidates for eventual reimbursement.
Public conversation and transparency
Thordarson discussed the project in a thread on X Saturday, describing how these tools could “democratize” the process of designing cancer vaccines-while emphasizing that some tumors in Rosie have not responded and further follow‑up is needed. His posts also acknowledged that informal experiments, even when ethically reviewed, will shape public expectations about what AI and mRNA can deliver for both pets and people.
Reaction from the technology sector reflected both curiosity and caution. “This is what I mean when I say the world is going to get very weird, very soon,” he wrote. “Expect more stories like this, each sounding increasingly more insane.” Matt Shumer, cofounder and CEO of OthersideAI, took to X to highlight the case, framing it as an early signal of how quickly individuals may begin orchestrating complex biomedical interventions with off‑the‑shelf AI tools.
What this does-and does not-tell us
- Signal, not proof: A single-animal response is encouraging but cannot establish safety, efficacy, or durability. For regulators and clinicians, Rosie’s case is a datapoint, not a precedent.
- Path to evidence: Carefully designed veterinary trials could clarify which cancers, targets, and dosing schedules matter-data that also inform human oncology and help agencies decide when AI‑assisted personalization crosses the line into standard care.
- Guardrails first: Any broader use must run through ethics review, quality-controlled manufacture, and appropriate regulatory pathways. Without that, the same tools that enabled one dog’s experimental treatment could fuel a parallel market of unregulated, AI‑designed biologics.
