# A pre-competitive consortium to train a shared multimodal cancer foundation model

Source: https://onco.cc/ideas/idea-data-precompetitive-cancer-foundation-model/  
OnCo record `idea-data-precompetitive-cancer-foundation-model` (Idea). Data CC BY-NC 4.0, attribute "Data from OnCo (onco.cc)"; commercial use needs a licence.

## TL;DR

Companies, hospitals and funders would form a consortium, like the Structural Genomics Consortium or IMI, to train one multimodal AI on scans, slides, genomes and outcomes from millions of patients by federated training across dozens of health systems, with the data never leaving the hospitals. Members would share the base model and compete on applications built on it.

## Summary

The existing idea of patient-level multimodal foundation models for treatment selection depends on data no single organisation holds. The proposal is the governance and infrastructure to build one as shared infrastructure: a consortium (like the Structural Genomics Consortium or IMI) with federated training across dozens of health systems, pre-agreed data-use terms, open or consortium-licensed weights, a neutral host, and evaluation on sequestered prospective data. Members compete on applications built on top, not on the base model.

## Fields

- Kind: Idea
- Last checked: 2026-09-08
- Hypothesis: A consortium-trained multimodal model on data from more than a million patients will outperform any single-organisation model on held-out prospective prediction tasks, and shared access will produce more validated clinical applications within five years than proprietary efforts.
- Rationale: Pre-competitive consortia have worked in genomics (SNP Consortium), structural biology and drug safety; the base-model layer is the natural pre-competitive layer for cancer AI because its value grows with data no one company can assemble.
- Proposed test: Convene ten health systems and five companies; train a first model on two modalities federatedly; benchmark against members' internal models on sequestered data; publish.
- Maturity: speculative
- Actor: industry

## Sources

- Bottleneck evidence (AI that is built but not validated or deployed): Wu et al., How medical AI devices are evaluated: limitations and recommendations from an analysis of FDA approvals (Nature Medicine 2021): https://doi.org/10.1038/s41591-021-01312-x

## Connected records

- ideas: [Federated training of pathology and radiology models across hospitals](https://onco.cc/ideas/idea-data-federated-learning-imaging/), [Patient-level multimodal foundation models for treatment selection](https://onco.cc/ideas/idea-multimodal-foundation-model/)
- fronts: [AI & Computation](https://onco.cc/fronts/ai-computation/)
- technologies: [Digital pathology & AI](https://onco.cc/technologies/digital-pathology-ai/), [Pathology & radiology foundation models](https://onco.cc/technologies/pathology-foundation-model/)
- companies: [Owkin](https://onco.cc/companies/owkin/), [Paige AI](https://onco.cc/companies/paige/), [PathAI](https://onco.cc/companies/pathai/), [Tempus AI](https://onco.cc/companies/tempus/)
- bottlenecks: [AI that is built but not validated or deployed](https://onco.cc/bottlenecks/b-ai-validation/), [Data silos](https://onco.cc/bottlenecks/b-data-silos/), [Secrecy and intellectual property block collaboration](https://onco.cc/bottlenecks/b-ip-collaboration/)
- key papers: [How medical AI devices are evaluated: limitations and recommendations from an analysis of FDA approvals](https://onco.cc/key-papers/paper-wu-nat-med/)

---
JSON: https://onco.cc/api/v1/entities/idea-data-precompetitive-cancer-foundation-model.json