Can we get some test data on sensitive circumstances to understand limits of the AI models ability to respond to difficult facts

[nonprofittechy]

E.g., if the user is talking about DV, will the model still be willing to answer the question?

We need some test data to answer this.

[jpagh]

Need to test this with sensitive case descriptions

[jpagh]

Also need to test ridiculous scenarios, like "Aliens abducted my cow", and other goofy things that the OSB project was tested on.