Getting useful content filtering result for troubleshooting in Azure OpenAI

Question

Hello there,

I'm calling Azure Open AI from Power Automate and want to see what's causing the trouble when prompt completion is stopped by (none other than) the content filter. However the returned body is not showing anything useful. Below is a sample of a blocked prompt. The 'content_filter_results' section is the same as a successful one (all 'safe'), while 'prompt_filter_results' indeed is only found in unsuccessful requests - but again it looks all normal (filtered = false, severity = safe, detected = false). Is there anyway to make get more info from the returned responses for troubleshooting?

"body": {
        "choices": [
            {
                "content_filter_results": {
                    "hate": {
                        "filtered": false,
                        "severity": "safe"
                    },
                    "self_harm": {
                        "filtered": false,
                        "severity": "safe"
                    },
                    "sexual": {
                        "filtered": false,
                        "severity": "safe"
                    },
                    "violence": {
                        "filtered": false,
                        "severity": "safe"
                    }
                },
                "finish_reason": "content_filter",
                "index": 0,
                "logprobs": null,
                "message": {
                    "role": "assistant"
                }
            }
        ],
        "created": 1725333306,
        "id": "chatcmpl-A3EIUYJ1OzMRICPnE8oZfGibYRpUh",
        "model": "gpt-4o-mini",
        "object": "chat.completion",
        "prompt_filter_results": [
            {
                "prompt_index": 0,
                "content_filter_results": {
                    "hate": {
                        "filtered": false,
                        "severity": "safe"
                    },
                    "jailbreak": {
                        "filtered": false,
                        "detected": false
                    },
                    "self_harm": {
                        "filtered": false,
                        "severity": "safe"
                    },
                    "sexual": {
                        "filtered": false,
                        "severity": "safe"
                    },
                    "violence": {
                        "filtered": false,
                        "severity": "safe"
                    }
                }
            }
        ],

Cheers,

GDB

Answer

Hi @romungi-MSFT , yes tried the interactive OpenAI studio and it's showing sth similar. Maybe that's the most OpenAI will provide? Wanna know at least the category triggering the filter. User's image

	{
		"role": "assistant",
		"content": [
			{
				"type": "text",
				"text": "

**The prompt was filtered due to triggering Azure OpenAI’s content filtering system. Please modify your prompt and retry.** [Learn more](https://go.microsoft.com/fwlink/?linkid=2198766)"
			}
		]
	}

Share via

Getting useful content filtering result for troubleshooting in Azure OpenAI

1 answer

Your answer