# A mandatory silent (shadow) trial before any cancer AI goes live

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

## TL;DR

Before an AI tool is allowed to influence care at a hospital, it would run invisibly alongside clinicians for months so its real-world performance at that site is known first.

## Summary

Shadow deployment (the model runs on live data but its outputs are hidden and compared with clinicians and outcomes) is standard practice at a few pioneering centres but not required. The proposal makes a pre-specified silent trial (minimum case numbers, pre-declared performance thresholds, subgroup analysis, comparison with local clinicians) a condition of go-live at each site, reported to the registry, as the AI equivalent of laboratory method verification before a new assay is used.

## Fields

- Kind: Idea
- Last checked: 2026-09-08
- Hypothesis: Site-level silent trials will identify locally unacceptable performance in a meaningful share of deployments that passed regulatory clearance, preventing harm and building trust at sites where the tool passes.
- Rationale: Clinical laboratories must verify every new assay locally before use, because performance depends on local conditions; AI has the same dependence and no equivalent rule.
- Proposed test: Require silent trials for all AI deployments in one hospital network for two years; report the proportion failing local thresholds and the reasons.
- Maturity: early-clinical
- Actor: clinic

## 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: [A standard for monitoring AI performance drift with pause thresholds](https://onco.cc/ideas/idea-data-drift-monitoring-standard/)
- fronts: [AI & Computation](https://onco.cc/fronts/ai-computation/)
- bottlenecks: [AI that is built but not validated or deployed](https://onco.cc/bottlenecks/b-ai-validation/)
- 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-silent-trial-before-deployment.json