> We reviewed a demonstration of this specific technique being used to identify a small number of previously known, minor vulnerabilities.
"small number", "previously known"+"minor"... they are trying hard to characterize this as harmless.
> These vulnerabilities all appear relatively simple, and we have found that other publicly-available models are able to discover them as well without requiring a bypass.
Ah so now they are admitting that this is all about hype after all.