My guess is that OpenAI are risk averse on this particular issue, because people could come up with some very creative awful scenarios with ChatGPT and image analysis. "Which of these people looks the most stupid" plus a jailbreak would be instant negative press.
On an unrelated note, I like your blog.