Alan Chan and Sam Manning, co-authors with OpenAI’s and Anthropic’s top researchers, say that many times, “internal safeguards have not been deployed.”