Gemini-2.5-flash returning finishReason: 'OTHER' and no explanation or candidates

Gemini API
, ,
1

I am making a request to Gemini API through Insomnia and I am getting this response:

{
	"candidates": [
		{
			"finishReason": "OTHER",
			"index": 0
		}
	],
	"usageMetadata": {
		"promptTokenCount": 259,
		"totalTokenCount": 259,
		"promptTokensDetails": [
			{
				"modality": "TEXT",
				"tokenCount": 1
			},
			{
				"modality": "IMAGE",
				"tokenCount": 258
			}
		]
	},
	"modelVersion": "gemini-2.5-flash",
	"responseId": "v-_CaK2zLJesz7IPt7GQyQE"
}

I am making a request to this URL: https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash:generateContent

With this payload (I removed the image base64 because I cannot share the image as it contains sensitive data):

{
   "contents":[
      {
         "role":"user",
         "parts":[
            {
               "inlineData":{
                  "mimeType":"image/jpeg",
                  "data":"[IMAGE_BASE64]"
               }
            }
         ]
      }
   ],
   "safetySettings":[
      {
         "category":"HARM_CATEGORY_HARASSMENT",
         "threshold":"OFF"
      },
      {
         "category":"HARM_CATEGORY_HATE_SPEECH",
         "threshold":"OFF"
      },
      {
         "category":"HARM_CATEGORY_SEXUALLY_EXPLICIT",
         "threshold":"OFF"
      },
      {
         "category":"HARM_CATEGORY_DANGEROUS_CONTENT",
         "threshold":"OFF"
      },
      {
         "category":"HARM_CATEGORY_CIVIC_INTEGRITY",
         "threshold":"OFF"
      }
   ],
   "generationConfig":{
      "temperature":0,
      "topP":1,
      "maxOutputTokens":872,
      "thinkingConfig":{
         "thinkingBudget":0
      }
   }
}

The problem seems to be the image because when I remove it from the request it works. If I paint the sensitive information from the image and send the new version the request also works.
The image is basically a product order containing a list of products and sensitive information such as address, name, telephone, tax number, brand name, brand website, order number, etc

3

Our content moderation pipeline includes configurable and non-configurable safety filters. The non-configurable filters, which are enforced regardless of user settings, automatically block certain inputs. For example inputs containing PII, references to private individuals, or other sensitive data.

The image you are trying to pass contains PII, which is causing this issue. For more details, we recommend reviewing the safety and content filter documentation.