So far, OpenAI, anthropic et al hasn’t sued anyone over it, but they have cut account access when it’s discovered to be used for that purpose

It’s how early versions of deepseek were trained iirc, it’s called distillation