“OpenAI and Anthropic are voluntarily increasing transparency about their AI model specifications, though not yet completely.”