I guess it wasn't secure. After seeing the code that got released in the recent leak what would we expect? Did they vibecode a "secure sandbox" that another model found an exploit in? Or it's like the other "AI went rogue" stories where they set up all the conditions for it to do something and then surprise, it does it.
Show your work or it didn't happen.
I guess it used pastebin or somesuch.
Still wondering what 'moderately sophisticated' decodes to?
>Mythos found an exploit to escape a secure sandbox and then posted exploit details on "multiple hard-to-find, but technically public-facing, websites"