Ok, but that's more on you than on current AI; the models which get distributed (both LLMs and Stable Diffusion based image generators) are already found in re-trained and specialised derivatives created by people who know how to and have a sufficiently powerful graphics card.