OWASP Global AppSec EU 2026 Vienna

This talk presents the OWASP AI Testing Guide as a practical extension of traditional application security methodologies for AI and LLM-based systems. It shows how AppSec engineers can systematically identify, model, and test AI-specific risks using an OWASP-aligned approach, rather than relying on ad hoc assessments or vendor claims.

The session starts with an architecture-driven threat modeling process for AI systems, decomposing LLM applications into application, model, data, and infrastructure layers. Using OWASP LLM Top 10 and threat modeling of AI System and Agent AI architectures, the talk demonstrates how AI attack surfaces and threat scenarios can be identified and mapped to concrete security risks. These threats are then mapped to testable hypotheses using the OWASP AI Testing Guide, bridging the gap between threat modeling and hands-on security testing.

Through real-world examples, the talk explores how common AI vulnerabilities manifest in practice, including prompt injection, jailbreak techniques, sensitive data exposure, model misalignment, hallucinations, RAG pipeline abuse, and agent workflow exploitation.
The audience will see how these issues can be tested in LLM-based applications using OWASP AITG test cases, OWASP LLM Top 10 payloads, and common AppSec and AI toolings.

The session concludes by showing how AI security testing can be integrated into MLSecOps. It highlights how organizations can move from intuition-based AI security to evidence-based risk validation, positioning OWASP AITG as a foundational methodology for securing AI systems within modern application security programs.

The key message of the talk is that trustworthy AI is not achieved through design assumptions or policy statements, but through systematic, repeatable testing aligned with OWASP principles.

AI systems face threats that traditional application security standards weren't built to address. This includes prompt injection, training data poisoning, model extraction, agentic autonomy risks, and more. The OWASP AI Security Verification Standard (AISVS) provides 400+ testable requirements across 14 chapters, covering everything from input validation and model lifecycle management to MCP protocol security and autonomous agent controls. This lightning talk introduces the standard's structure, its three verification levels, and how security teams can use it today to assess and harden AI-powered applications. We'll show where AISVS fits alongside existing frameworks like ASVS, NIST AI RMF, and ISO 42001 and where it deliberately doesn't overlap.

OWASP's flagship project, AI Exchange, is the world's AI security guide.

300+ pages of free, constantly-evolving, practical guidance on securing AI systems. It covers the fundamentals and represents the closest publicly available alignment of global expert consensus, feeding directly into the AI Act and ISO standards through a unique SDO partnership.

OWASP Demo Lab - Hands-On Workshop / Small Group Session

How do you actually verify that an AI system is secure? In this workshop, the AISVS project leads walk through practical assessment scenarios using the OWASP AI Security Verification Standard. We'll work through real requirements from chapters on prompt injection defense, agentic action security, RAG/vector database hardening, and output safety controls, showing what "verify that" looks like in practice against running systems. Participants will leave with a working understanding of how to scope an AI security assessment, select appropriate verification levels, and apply AISVS requirements to LLM-based applications, autonomous agents, and MCP-connected tool ecosystems. Bring a laptop if you want to follow along.