Generative artificial intelligence systems, such as OpenAI ChatGPT and Anthropic Claude, have shown remarkable capabilities in generating human-like text responses, raising important ethical questions about their behavior when interacting with controversial topics. The novel concept of systematically measuring hostility in AI responses to maliciously crafted prompts reveals significant differences in the behavior of LLMs, highlighting the importance of developing robust and ethically aligned AI systems. The study employs a comprehensive methodology involving prompt engineering, automated response analysis, and quantitative metrics to evaluate the hostility levels of AI-generated content. Findings indicate that ChatGPT exhibits higher hostility metrics compared to Claude, showing the need for enhanced training protocols and ethical guidelines. The implications for AI development and deployment are discussed, emphasizing the necessity for continuous monitoring and interdisciplinary collaboration to ensure the safe and responsible use of AI technologies. Recommendations for mitigating hostility in AI responses are provided, contributing to the broader effort to create trustworthy and reliable AI systems.