AI results can depend on input so consider how to use these ideas from Stanley Lieber in forging better queries ...
When combined with another exploit, the "PromptFiction" vulnerability, which has been fixed, could have enabled an end-to-end attack on a targeted system.
Claude for Chrome v1.0.80 still accepts forged clicks from other extensions, triggering Gmail, Docs, and Calendar tasks ...
A complex AI safety prompt can contain rules that conflict under specific conditions. Those conflicts are easy to write and hard to see by eye. rule-audit reads the prompt the way a linter reads code ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results