Open AI and Anthropic have already begun restricting reverse engineering with their models.