OWASP GenAI Security Training in Kraków, Poland

This training provides security professionals, developers, and AI practitioners with a structured understanding of the OWASP Top 10 for Large Language Model (LLM) Applications, GenAI risks, threat modeling, secure development practices, and mitigation techniques for AI-driven systems.
The course blends theoretical knowledge with hands-on demonstrations involving prompt attacks, model abuse scenarios, and defensive controls.

Foundations of GenAI Security & OWASP LLM Top 10

1. Introduction to GenAI & Security Basics

• What is Generative AI (GenAI)?
• How GenAI apps differ from traditional applications
• Key components: LLMs, prompts, embeddings, vector DBs
• Why GenAI security matters

2. Introduction to OWASP GenAI / LLM Top 10

• Purpose of OWASP GenAI Security Project
• How GenAI risks map to traditional security risks
• Overview of all 10 categories with simple real-life examples

3. Deep Dive into OWASP LLM Top 10 (Part 1)

• Prompt Injection
• Insecure Output Handling
• Training Data Poisoning
• Model Theft / Extraction
• Supply Chain Vulnerabilities

4. Deep Dive into OWASP LLM Top 10 (Part 2)

• Data Leakage
• Unauthorized Code Execution
• Hallucinations & Overreliance
• Excessive Agency
• Insecure Plugin / Tool Integration

5. Hands-On Demo

• Simple prompt injection examples
• Manipulating instructions
• How insecure outputs lead to risk

Threat Modeling, Secure Development & Mitigation

6. Threat Modeling for GenAI

• Understanding data flow in LLM-based systems
• Finding attack points
• STRIDE basics applied to GenAI
• Misuse / abuse scenarios

7. GenAI Architecture & Common Risks

• API calls & model gateways
• Risks in hosted vs self-hosted models
• RAG (Retrieval-Augmented Generation) risks explained simply
• Risks in plugins, tools & AI agents

8. Secure Development Practices

A. Secure Prompt Engineering

• Designing safe prompts
• Guardrails & boundaries
• How to reduce prompt injection
• Output validation basics

B. Data Security & Privacy

• Safe data ingestion
• Masking & minimization
• Handling sensitive data

C. Supply Chain Security

• Risks in pretrained models
• Risks in libraries & open-source packages
• Model integrity checks

9. Defensive Controls & Testing

• Input/output validation
• AI safety tools (guardrail models, AI firewalls)
• API security basics (rate limits, authentication)
• Testing for:
• Hallucinations
• Unsafe outputs
• Prompt injection

10. Governance, Compliance & Best Practices

• Responsible AI basics
• Key regulatory frameworks (simple explanation)
• Documentation & audit readiness

11. Final Lab + Q&A

• Small hands-on scenario
• Review of OWASP LLM Top 10
• Best practices checklist