GPT model switching and lock to "restricted version of GPT‑5" without consent?

ChatGPT
1

Hi everyone,
I’m a Plus subscriber and I’ve noticed something odd over the past few days. Even when I select GPT‑4o (or even normal GPT-5 Model), after a while of discussion with GPT (usually when we get past the temporary context window), The model switches to an heavily censored inferior version of GPT-5 without my consent and stay locks on it, no matter how much i try to change it or what i talk about it will lock on this version of GPT-5 which wont follow any of prompts and only gives basic answers, like even when i use gpt camera to take picture it tells me that it cannot talk about my look!
Support told me this is part of a “safety routing” test that automatically moves certain messages to GPT‑5 when the system detects emotional or sensitive content, but im suspecting its more like a problem/bug which they are not aware of.

This means that:

  • My chosen model isn’t always the one actually responding.

  • I can’t lock a chat to GPT‑4o, even though that’s what I’m paying for.

  • Im forced to use a routed model which behaves more limited and filtered, which affects ongoing work and projects.

I’d like to know:

  • Are other Plus users seeing this behaviour too?

  • Does anyone know if OpenAI plans to add an opt‑out or model‑lock feature?

  • How do you feel about messages being rerouted between models without notice?

I’m sharing this so users can compare experiences and push for more transparency. We all deserve to know exactly which model we’re using and to have control over it.

Thanks for any insights or confirmations from your side.

1 Like
2

Hell a moment ago i gave it a IMAGINARY company political scenario (which didnt broke any policy or law) and asked it to which provided option is more likely the problem and the model was immediately switched to limited GPT5 which gave me a stupid answer of it cannot talk about that or this and i was not able to change the model or get any other answer!
This is beyond even normal censoring! it is bland limiting user to any personal view Openai has about anything!

3

This topic was automatically closed after 24 hours. New replies are no longer allowed.