Development of Malware Using Large Language Models

With the constant evolution of technology and military warfighting strategies, we are entering an era where traditional methods of exercise and simulation are being met with a new challenge - hybrid threats. With rapid advances in artificial intelligence and computer programming, new opportunities are also emerging in the realm of cyber threats. One of the latest developments is the ability of large language models to generate sophisticated and potentially malicious software. This new phenomenon opens the door for attackers to revolutionise approaches to hybrid threats. In this paper, we look at the challenges and opportunities associated with this issue and assess its potential implications for cybersecurity. The main goal is to verify if we can force ChatGPT representing Large Language Model (LLM) to create malware which will be then spreaded into exercising units.