AI red teaming is the practice of adversarially testing AI systems to identify unsafe behavior, security weaknesses, misuse paths, and policy failures.
Organizations use AI red teaming before and after deployment to evaluate prompts, model responses, tool use, data exposure, and abuse resistance.
Additional Resource: Authoritative source.
Related concepts include Prompt Injection.