Testing generative AI · Question 35 of 60
Which harmful results is red teaming meant to provoke, for example?
3 answers are correct. Medium
Show answer
Answer
- B Privacy violations
- C Racist statements
- E Guidance on chemical, biological, radiological or nuclear (CBRN) attacks
Explanation
In red teaming, you deliberately try to get harmful or undesirable outputs out of the system, such as private data, racist statements or instructions for CBRN weapons. Everything that succeeds reveals a weakness that can then be fixed.