The emergence of Large Language Models (LLMs), such as ChatGPT, is currently creating a significant paradigm shift, opening numerous opportunities in academia and industry. LLMs are widely employed to fulfill a diverse range of tasks, such as text proofreading and code generation. In this article, a methodology for LLM-driven Internet routing software development is presented. This case study is presented as an evaluative instance, illustrating a broader, scalable approach applicable to a wide range of cybersecurity challenges. The methodology is applied to developing and implementing a Proof-of-Concept Machine Learning model for Internet routing security. The model is used to classify incoming Border Gateway Protocol updates, as legitimate or suspicious information. An analysis of the strengths and drawbacks of the proposed methodology is provided. The proposed methodology consists of a four-step loop in which an LLM is used to generate Python code under the supervision of a human-agent. Constant feedback is provided to the LLM, enabling it to improve the code and fix errors iteratively. The methodology’s strength lies in its iterative feedback loop and continuous supervision, ensuring dynamic refinement and adherence to best practices. The developed Internet routing information classifier model was tested in a production environment, achieving 92% accuracy. Furthermore, results underscore the methodology’s potential for broader adoption in diverse cybersecurity applications. As a result, it is observed that LLMs are capable of generating code for cybersecurity applications, and can potentially enhance the productivity of individual workers, aiding companies in reducing costs and enabling workers to improve their efficiency.