Remote AI Adversarial Red Team Specialist
Summary
You probe AI models with adversarial inputs to find vulnerabilities, classify failures, and document reproducible attack cases for customers to improve safety.
Mercor is building a red team for AI safety, probing models with adversarial inputs to surface vulnerabilities and generate data that helps customers improve their systems. The project involves reviewing AI outputs that touch on sensitive topics; participation in higher-sensitivity projects is optional and guided by clear guidelines and wellness resources.
You will annotate failures, classify vulnerabilities, and document reproducibly with reports, datasets, and attack cases for customers to act