Couldn’t this potentially be just another completion, where the AI is “playing along” and generating responses as if the user had broken through, even though the user had not?
How would we even know?
How would we even know?
> On Thursday, a university student named Marvin von Hagen independently confirmed that the list of prompts Liu obtained was not a hallucination by obtaining it through a different prompt injection method: by posing as a developer at OpenAI.
Although one cannot be 100% certain.
I don't know if it was a hallucination or not, but what I've been given here doesn't rule the idea out for me, and in fact the lack of sophistication around the reporting and "research" leads me to believe the idea wasn't even considered seriously.
Horrible reporting, this is a nothing burger, and Marvin von Hagen has next to no background in AI or this technology and zero authority to claim whether or not it's a hallucination.