Azure AI Content Safety: Prompt Shields vs blocklist vs Task Adherence vs custom categories vs multimodal

Verdict: Hidden instructions in documents need Prompt Shields. Fixed banned terms need a blocklist. Misaligned tool calls need Task Adherence. Emerging brand patterns need custom categories. Image plus caption needs the multimodal API.

CriterionPrompt ShieldsCustom blocklistTask AdherenceCustom categories rapidMultimodal analyze API
Detects or doesInstructions hidden in user prompts or third-party documentsBlocks a fixed set of terms regardless of severity scoreFlags agent tool calls or responses misaligned with the user taskScans text and images for emerging brand-specific patternsScores an image together with its caption in one call
Choose whenA RAG or agent pipeline ingests third-party or OCR'd document textReject specific competitor or internal code namesAn agent invokes tools with inputs not matching intentBrand-specific prohibited symbolsCombined image-plus-caption moderation

Rules

Traps