The integration of artificial intelligence into commercial applications has revolutionized various industries, offering unprecedented capabilities and efficiencies. However, the vulnerabilities associated with large language models, particularly in the context of prompt extraction threats, present significant security challenges that must be addressed to protect sensitive information and maintain user trust. This research evaluates prompt extraction vulnerabilities in ChatGPT, focusing on its deployment across customer service, content creation, and educational tools. An automated threat simulation was employed to systematically test the model's response to crafted prompts, and the results were analyzed using quantitative metrics to assess the severity and success of information extraction. The study's findings revealed substantial risks, with the highest extraction success rates observed in educational and content creation contexts. The implications for data privacy and model security are profound, emphasizing the need for advanced defensive strategies and continuous improvement in LLM deployment. This research offers a structured framework for evaluating LLM security and provides actionable insights for developing more resilient models, ultimately contributing to safer and more reliable AI applications.