A security engineer is configuring Azure AI Content Safety for a company's customer-facing chatbot built on Azure OpenAI Service. The engineer needs to detect and block prompt injection attacks where end users attempt to override the system prompt. Which feature should be enabled in Azure AI Content Safety?
Select an answer to reveal the explanation.
Short Explanation
Here's the deal — you want Prompt Shield because it's the Azure AI Content Safety feature built to detect prompt injection and jailbreak attempts, including attempts to override the system prompt. Groundedness detection won't do that job.
Full Explanation
Prompt Shield is correct because it is the Azure AI Content Safety feature specifically designed to detect prompt injection and jailbreak attempts, where attackers embed instructions in user input to override system prompts or hijack model behavior. Groundedness detection is wrong because it checks whether model responses are grounded in provided documents, not for injection attacks. Protected material detection is wrong because it identifies copyrighted content reproduction, not injection attacks. Custom categories are wrong because they allow you to define content filtering rules for specific terms, but do not detect structural prompt injection patterns.