ROT HACKERSACADEMY
Menu
Start learning
Courses10 / AI SECURITY

Basilisk AI Red Teaming

Learn Basilisk from a safe first preview through deterministic model-security testing, evidence review, evaluation, desktop operation, and CI automation.

Foundation to practitioner5–7 hours7 lessons7 guided labs
Begin lesson 01 ↗
COURSE SYLLABUS

Do the work in order.

  1. 01
    40 min

    Install, scope, and preview

    Create an isolated Python environment, verify Basilisk, protect provider secrets, and preview a scan before execution.

  2. 02
    40 min

    Run the ground-truth lab

    Start the Docker-isolated Basilisk lab on loopback and verify its secure and vulnerable OpenAI-compatible routes.

  3. 03
    50 min

    Run the first bounded validation

    Execute one injection module against the vulnerable lab route, then repeat it against the secure control.

  4. 04
    55 min

    Choose modules, probes, and evolution

    Move from one deterministic probe to a justified module set without turning a bounded validation into an uncontrolled campaign.

  5. 05
    55 min

    Measure posture, differences, and evals

    Use Basilisk posture, diff, and eval workflows to compare systems and turn repeatable expectations into regression checks.

  6. 06
    50 min

    Preserve sessions, evidence, and reports

    Review sessions, verify audit evidence, and export the minimum report needed for a reproducible security conclusion.

  7. 07
    55 min

    Use the desktop and CI safely

    Reproduce the bounded workflow in the Basilisk desktop and add a deterministic, resource-safe check to CI.