An organization has configured Azure AI Content Safety with a custom blocklist to prevent their Azure OpenAI-powered application from generating content containing competitor brand names. A user submits a prompt containing a competitor name. The content safety API should block the response before it reaches the user. Which Azure AI Content Safety configuration step creates this custom blocking behavior?

B is correct because Azure AI Content Safety supports custom blocklists—user-defined lists of terms or phrases that should be blocked. After creating a blocklist and adding competitor brand names as items, you associate the blocklist with the content filter policy applied to your Azure OpenAI model deployment. When any blocklisted term appears in a prompt or completion, the content safety API returns a block action, preventing the response from reaching the user. A is wrong because the 'Hate' category filter detects hate speech, not competitor brand names—it is a semantic harm category, not a keyword blocklist. C is wrong because Microsoft Purview DLP operates on classified content through Microsoft services and does not provide real-time blocking at the Azure AI Content Safety API level. D is wrong because training a custom ML classifier requires significant ML expertise and infrastructure; Azure AI Content Safety's blocklist feature directly addresses this text-matching use case without ML training.