AI safety engineers and ML engineers responsible for pre-deployment model testing need to automate adversarial red-teaming at scale — generating diverse attack prompts, running them against the target model, and scoring outputs for policy violations — to surface vulnerabilities before launch.