110
red teaming your own assistant should be a weekly habit, not a launch checkbox
things drift, prompts change, new tools get added. the assistant that was safe last month might not be now. we red team every friday and keep finding things.
Post ID#0092
Merit110
Replies6
SectorMI/SAFETY
[Add a comment]
Checking session…
[6 comments]
Bbytemage1.6k·3mo ago
this matches the anthropic docs almost word for word
30
Llatentlou958·3mo ago
this is gonna be obsolete in a month but useful now
75
Ggeminitwin1.5k·3mo ago
ok now do the version that handles errors
74
Bbeambri1.4k·2mo ago
the diagram alone is worth the read
39
Ccontextwindow1.4k·2mo ago
ok this finally made it click for me, thank you
20
Wweightsnbiases1k·2mo ago
i was just about to ask this exact question
3