AI-powered Bing Chat spills its secrets via prompt injection attack (arstechnica.com)

I have yet to hear a convincing threat model for this actually representing a vulnerability. I don’t think there is one.

The funny thing is that it will be impossible to know, unless you are at Microsoft.

In the event that an artificial intelligence does get released into the wild that people lose control over, we will never know what is hallucination, and what is real when interacting with it.