Azure GPT endpoints with vision capabilities have faces in the image blurred before processing

Question

Hello! If I send an image to the model and ask it to detect what expression the face in the picture is making, the model won't comment on the expression, bur rather the posture and other details of the image. If I explicitly ask the model whether the face in the image is blurry, it says it is.

This happens with both gpt-4o-mini and gpt-4.1-mini. Note that I tried the same exact models in the OpenAI API and those are perfectly capable of describing the expression in the image I send them. I assume the reason behind this is privacy related. I wonder if it would be possible to turn off this behavior somehow?

Answer 1

Hi @Nicola De Angeli,

The behaviour you're noticing is expected and comes from a privacy feature built into Azure OpenAI’s vision capabilities. According to Microsoft’s policy, face blurring is applied as a preprocessing step, which means any faces detected in an image are blurred before the image is passed to the model for analysis. This process ensures that the model cannot uniquely identify individuals and helps Microsoft comply with privacy regulations and responsible AI principles.

Because of this, the model is unable to recognize or describe facial expressions, as it only sees a redacted version of the image. Even if you directly ask about the face, the model will often respond that it appears blurry, which is accurate given what it's receiving. This behaviour is enforced at the platform level and cannot be disabled. Please look into to below screen-shot for more info.

For more info, please refer to the documentation: Azure OpenAI Transparency Note.

I hope you understand. And, if you have any further query do let us know.

---

If this answers your query, do click Accept Answer and Yes for was this answer helpful.