AI guardrails stripped from Meta and Google models in minutes (ft.com)

5 points by uxhacker 3mo ago ↗ HN

2 comments of 4

[ 21.9 ms ] story [ 16.9 ms ] thread

Is not the answer to strip the dangerous information before training? Rather than trying to add guardrails post training.

[dead] dividendflow 3mo ago ↗

[dead]