I ran 3,360 safety tests on GPT-4o, Claude, Grok, DeepSeek, Gemini (github.com) 4 points by aestrad7 3mo ago ↗ HN
[–] inaros 3mo ago ↗ Great work.TLDR: 42 attack types. 5 models. 3,360 tests. 1 in 3 harmful requests got through.
[–] AlejaGiral31 3mo ago ↗ Wow! Excellent independent work! I can't believe Grok performed so well! How did you ensure all models were tested equally?
3 comments
[ 4.2 ms ] story [ 13.2 ms ] threadTLDR: 42 attack types. 5 models. 3,360 tests. 1 in 3 harmful requests got through.