ReadyNow.AI

AI Red-Teaming with Viper:

• ReadyNow.AI — Red-Teaming AI Systems with Viper
  In just one day, use the Viper toolkit to attack, fix, and report on real AI systems—no AI background required. Leave with proven tools, evidence, and management-ready templates to kickstart your organization’s AI security journey.
  
  Duration: 1 Day (09:00–17:00)
  Target Group: Cybersecurity professionals with little or no AI/ML experience
  Format: Instructor-led, step-by-step labs, all environments provided
  Minimum Requirements for Participants
  Technical Background
  Basic familiarity with cybersecurity concepts (e.g., penetration testing, vulnerabilities, incident response)
  Ability to read and run simple Python scripts (no development or data science experience required)
  Comfortable using terminal/command line for basic operations
  Hardware
  Laptop (Windows, macOS, or Linux) with at least 8 GB RAM and at least 20 GB free disk space
  Reliable Wi-Fi/internet connection
  Software
  Modern web browser (Chrome, Firefox, Safari, or Edge)
  Docker Desktop installed and running (mandatory for lab exercises)
  Git installed (recommended for cloning repositories)
  Ability to install Python packages if required (all scripts and containers provided)
  Other
  Administrative rights on the laptop to install or run Docker/Git (no exceptions)
  No prior AI or machine learning experience necessary
  All code, target apps, attack recipes, reporting templates, and sample data will be provided during the course
  Course Outcome
  Participants will use the Viper toolkit to attack a realistic AI system, identify vulnerabilities, apply fast mitigations, and generate a professional evidence/report pack—using clear, repeatable “recipes.”
  
  Course Structure
  
  Morning Block (09:00–12:30)
  Viper Setup & Core Attack Techniques
  Introduction & Risk Framing
  Overview of generative AI vulnerabilities in simple terms
  Quick demo of a live attack using Viper
  Lab: Environment Setup
  Launch Viper in Docker (step-by-step or pre-provisioned)
  Run a provided vulnerable app (chatbot or RAG-based demo)
  Explore Viper’s interface and attack modules
  Lab: Prompt Injection & Jailbreaks
  Use built-in attack recipes for prompt manipulation and filter bypass
  Capture logs, screenshots, and failed responses
  Lab: Data Leakage & Indirect Access
  Trigger extraction of hidden or sensitive data
  Explore attacks via poisoned content or document injection
  Document findings for reporting
  
  Afternoon Block (13:30–17:00)
  Mitigation, Testing & Reporting
  Lab: Quick-Fix Defenses
  Apply prompt-level guardrails and allowlists using Viper’s mitigation examples
  Test effectiveness by re-running earlier attacks
  Lab: Logging & Monitoring
  Enable simple telemetry features to track live attacks
  Validate whether protections block or log threats
  Report Pack Assembly
  Export logs, screenshots, and fill in the provided reporting template
  Build a concise “Findings & Fixes” summary for security or management teams
  Debrief & Next Steps
  Where to find new attack modules and keep Viper updated
  How to use Viper in CI/CD or DevSecOps environments
  Open Q&A and workshop feedback
  
  Your Takeaways
  Solid working knowledge of modern AI system vulnerabilities—practical, not theoretical
  Ready-to-use Viper Docker lab (runs on any laptop)
  Step-by-step recipe notebook for real-world AI attacks & fixes
  Editable reporting templates (audit/CISO-ready)
  Evidence pack: logs, screenshots, attack/fix artifacts

Extension

Note: This course is also available as an extended 2-day version, with Day 1 dedicated to in-depth coverage of Model Security. This foundational track explores threats across the model lifecycle—from data poisoning and prompt injection to model theft and adversarial inputs—while introducing core mitigation strategies, security design patterns, and the evolving regulatory landscape. It provides a critical security lens that complements the governance structures developed on Day 2.