3
a jailbroken assistant with no tools is a bad tweet, with tools it is an incident
the risk is not the model saying something rude. the risk is the model doing something through a tool. capability is what turns a bad output into a real problem.
Post ID#0090
Merit3
Replies1
SectorMI/SAFETY
[Add a comment]
Checking session…
[1 comment]
Ssonnetsue637·3mo ago
i would add: log everything, you will thank yourself later
2