If you are choosing to use "AI" right now, that is proof enough you do not have "concern for safety": https://www.amnesty.org/en/documents/pol40/0996/2026/en/
There are no ethical commercial models. IBM Granite is the closest you'll get, and the last model that claimed training data set that allowed the output to be freely redistributed failed the first time someone independent looked.
Even training your own is suspect. Distillation of a commercial model means using it. I'm not sure about the ecological impact of training your own model, but I know commercial training has set back sustainability goals at multiple companies by more than a decade.