LLM-powered applications introduce a new class of risks - from prompt injection and data leakage to hallucinations, model abuse, and indirect exploit paths that traditional security models don’t cover. This session examines how these vulnerabilities emerge in real-world AI systems and how they can be systematically controlled.

Explore exploit patterns and safety architectures for LLM applications, and see how SAP AI Core enables secure-by-design adoption through seamless LLM integration, embedded guardrails, policy enforcement, and operational controls - helping organisations move from AI experimentation to production-grade, trustworthy systems.