Passed Persona verification – but access to gpt-image-1 API still denied due to internal error?

Hi all,

I’d like to report a very frustrating experience that I believe may be affecting other users as well — especially those who completed identity verification via Persona.

On June 10, I successfully passed identity verification using the official Persona flow. I received a clear success message
There was no indication of any rejection, failure, or document issue.

However, despite this, I still don’t have access to API features like gpt-image-1. The model appears in the list of available models when calling /v1/models, but I get the error:
“Access denied. Check API key and permissions.”

I contacted support multiple times. At first, I was told to wait 30 minutes, then a few hours, then up to 5 days. Finally, I received the following message from support:

“Even if the verification appeared completed on your side, it might have failed silently due to unclear documentation, unsupported ID, or mismatched information. Unfortunately, we can’t offer retries.”

This is extremely frustrating. I used a real passport. I even saved my personal verification link and took a screenshot of the success message. The issue most likely occurred due to OpenAI server problems on June 10 — the API status page showed significant downtime that day.

Yet, users like me are now being told that there is no way to retry, and we’re simply locked out from access — despite being willing to verify again and to pay for higher API usage than ChatGPT.

Why is this acceptable?

I’d like:

  1. A clear explanation of what caused this verification to silently fail.
  2. The ability to retry verification or have someone manually review my case.
  3. Clarification on why I can use gpt-image-1 within ChatGPT, but not via the API.

If others have experienced the same issue — please reply. This needs attention.

Thank you.


I passed verification via Persona (June 10), received the success message, but my verification_status for PERSONA_SELFIE_AND_ID remains "STARTED" — and I’m locked out of gpt-image-1 access due to a backend issue.

2 Likes

I sympathize with your case but I think the root mistake here was offering all this personal info to OpenAI in the first place. There is no legitimate reason to perform such invasive checks over an AI.

This would presumably give hints to whatever supposed bad guys OpenAI believes exist.

That’s the case for everyone.