extends classic penetration testing with security analyses specifically tailored to AI systems and large language models. Based on established standards such as the OWASP Top Ten for LLMs, vulnerabilities are identified that often remain undetected in conventional tests. The goal is to identify risks in the use of AI at an early stage and to increase the security of your applications in the long term. By combining AI-specific tests, in-depth analysis of security-relevant attack points, and clear recommendations for action, @-yet supports companies in sensibly expanding their existing pentest strategies and reliably securing AI systems.