I wanted to start a broader ideas worth trying thread around AI Red Teaming.
A useful angle is common mistakes, but it also connects with evaluation. Different tools and workflows can produce very different results even when the goal is similar.
It would be interesting to compare practical...