With the increasing use of AI systems, security risks are also growing – from manipulation and hallucinations to data leakage and misuse, as listed in the OWASP Top 10 for LLMs, for example. Addressing these risks requires a clear approach to AI security and governance. The F5 AI Gateway provides practical protection mechanisms for this purpose, including prompt security and guard rails to intercept risky inputs. In addition, features such as identity-aware model routing, language detection, and protection against information leakage support secure model usage. The presentation shows how AI applications can be operated in a reliable, compliant, and trustworthy manner.