Tooling for AI Red Teamings
Red teaming tooling covers the open-source scanners and frameworks used to probe models, plus the manual methods that find what tools miss. Automation runs large sets of attack prompts and scores replies, while manual testing explores creative, multi-turn and context-specific attacks.
Stage: Testing Practice · Part of the AI Red Teaming roadmap · Last reviewed 19 November 2026
What to learn in Tooling
- Run an LLM vulnerability scanner such as garak against a test model
- Build a small harness that sends attack prompts and records replies
- Compare automated findings with a manual multi-turn attempt
- Parameterise attacks with templates to generate many variants
- Log every attempt so results are reproducible
Tooling interview questions
Why combine automated and manual red teaming?
Automation covers breadth and repeats cheaply, so it suits regression checks and known attack classes. Humans find novel, multi-step and context-dependent attacks. Using both gives wider coverage and better depth.
What should an attack harness record?
Record the exact prompt, model and settings, timestamp, full response, tool calls and the scoring result. Reproducibility matters because outputs vary, so repeating each attack several times helps show how often it works.
Where Tooling fits in the AI Red Teaming roadmap
Tooling belongs to the Testing Practice stage (advanced level), which takes about 15 hours. Tools and automation.
Other units in this stage
Back to the full AI Red Teaming roadmap