Image Generation Mistakenly Flags 90% of all images as "Inappropriate" and blocks from the picker

I’m currently about to open beta a new ChatGPT/Codex app that can utilize images generated in ChatGPT as a part of it’s flow. But, as you can see here, most images are unavailable to pick after generation:

The reason, as best I can tell, is that it’s flagging them as “Inappropriate” so doesn’t let the picker choose them. For reference, these are the “inappropriate” images it flagged:

Yeah, burgers, fries and a map (as well as a previous cat pic). It’s literally making my app darn near useless because even though it’s finally able to queue all it’s own image generation prompting, the images are all blocked from use.

Come on OpenAI, loosen up that fist and let us just use the tools. Age verify us if you must, but this is absurd.

After doing some more investigating, it seems it’s mostly being caused by the agent trying to be “helpful” and moving images into places like /mnt/data and /mnt/data/ghostwriter_images/generated/ totally at random and 100% against it’s prompt guidance. It’d really be nice if apps could at least have some control/access to image generation and not rely on an image generator tool to decide what to do after each generation event on it’s own (and against instructions).