[OpenAI] also emphasized that the ChatGPT model implicated is no longer available and suggested that current models are safer and "The safeguards in ChatGPT today are designed to [...] safely handle harmful requests

These sentences are certainly meant as a defense, but to some they may also sound like an admission.

A more active thread: https://news.ycombinator.com/item?id=48110689