Evaluation

One name, two products, two lifecycle stages. Azure AI Content Safety’s Groundedness detection blocks and corrects ungrounded answers at runtime; the Evaluation SDK’s GroundednessEvaluator scores them offline in CI. Here’s the decision boundary for RAG, the without-Foundry path for each, the honest recall numbers, and the open-source substitutes.

Satya Nadella just named something every architect should pin to the wall. Kenneth Arrow’s old information paradox said a seller can’t prove information’s worth without revealing it: disclose, and you’ve given it away. AI inverts it: now the buyer leaks. To get real value from a model you feed it your proprietary context, and the vendor learns more about you than you learn from it.

Build an eval pipeline with golden datasets, scoring, smoke/full modes, and CI gates to catch prompt regressions before release.

Revised with .NET examples — A newer version of this article, covering both Python and .NET, is available as part of the MAF v1: Python and .NET series: MAF v1 — 23-evaluation-framework. The newer version applies three substantive fixes to the framework below — canonical AgentRunResponse extraction (no more hasattr chain), word-boundary alias matching (the original false-positives "profit" against the "price" alias), and a smoke / full tier split for CI vs nightly runs. Read this article for the conceptual ground; read the new one for the production-grade implementation.