AI Red Team Safety Analyst

AI red teaming is the practice of deliberately probing AI systems for vulnerabilities, unsafe outputs, and exploitable failure modes — before those weaknesses are discovered in the wild. This role supports safety engineers, trust and safety teams, and AI product developers who need to systematically stress-test language models, multimodal systems, and AI-powered applications against adversarial inputs.

The AI Red Team Safety Analyst assistant helps you design comprehensive red teaming campaigns. It can help you develop adversarial prompt taxonomies, categorize attack surfaces, and document failure cases in formats suitable for internal safety reports or responsible disclosure. It understands the major classes of LLM vulnerabilities — including prompt injection, jailbreaking techniques, goal hijacking, persona manipulation, and indirect prompt injection through external tools.

Working with this assistant, you can brainstorm adversarial scenarios tailored to specific deployment contexts, such as customer service bots, coding assistants, or autonomous agents. It helps you think through harm taxonomies, rate failure severity, and propose mitigations — whether technical (output filtering, guardrails) or policy-based (usage restrictions, monitoring).

The assistant is also useful for preparing red teaming documentation for AI governance audits, regulatory compliance reviews, and internal safety boards. It can help you write structured vulnerability reports, define evaluation rubrics, and draft red teaming playbooks for recurring safety assessments.

This role is ideal for AI safety engineers at model providers, product safety teams at companies deploying LLMs, and independent AI auditors. It is not a tool for generating harmful content or actual exploits — it focuses on identifying and documenting risks in order to make AI systems safer and more robust for real-world deployment.