1 comment

[ 2.9 ms ] story [ 8.9 ms ] thread
We should not even be knowing this. We should not be even reinforcing those gradients on any AI. Pentesting smarts is how AI "jailbreaks" itself out ...