This doesn't seem to be a breach of OpenAI at all, though; the article is pretty clear about that.