{"entity":{"id":"idea-data-regulatory-sandbox-adaptive-ai","kind":"idea","name":"A regulatory sandbox for continuously learning cancer AI","aka":[],"tldr":"Let AI tools that improve as they learn be used under close supervision in a few hospitals, with pre-agreed rules for what changes are allowed and how they are checked.","summary":"Regulation assumes a frozen model; models that update on new data are effectively unapprovable, so deployed models age. The FDA's predetermined change control plan is a step. The proposal is a formal sandbox: a small number of sites, a pre-specified envelope of allowed updates, mandatory shadow evaluation of each update on sequestered data before activation, full audit logs, and a regulator seat at the table, generating the evidence needed to write general rules for adaptive AI.","asOf":"2026-09-08","links":[{"label":"FDA Predetermined Change Control Plans (page moved; nearest live section)","url":"https://www.fda.gov/regulatory-information/search-fda-guidance-documents/"}],"tags":[],"related":[],"cancers":[],"sections":["ai-computation"],"technologies":[],"targets":[],"drugs":[],"companies":[],"institutions":[],"pathways":[],"terms":[],"trials":[],"people":[],"bottlenecks":["b-ai-validation","b-regulatory-fragmentation"],"keyPapers":[],"journals":[],"dependsOn":[],"notes":[],"hypothesis":"Adaptive models in the sandbox will maintain or improve performance over two years while frozen comparators degrade, without safety events attributable to updates, providing the evidence for a general adaptive-AI pathway.","rationale":"Sandboxes in fintech produced workable regulation for novel products faster than rulemaking; the same approach suits AI where the risks are poorly understood in advance.","test":"Run a two-year sandbox with three adaptive models at five sites; compare performance trajectories with frozen versions; publish the regulatory learnings.","maturity":"early-clinical","actor":"regulator","cost":"medium","horizonYears":3},"route":"/ideas/idea-data-regulatory-sandbox-adaptive-ai/","neighbours":{"section":[{"id":"ai-computation","kind":"section","name":"AI & Computation","route":"/fronts/ai-computation/"}],"bottleneck":[{"id":"b-ai-validation","kind":"bottleneck","name":"AI that is built but not validated or deployed","route":"/bottlenecks/b-ai-validation/"},{"id":"b-regulatory-fragmentation","kind":"bottleneck","name":"Regulatory divergence between regions","route":"/bottlenecks/b-regulatory-fragmentation/"}]}}