LLM-based systems are applications that leverage Large Language Models (LLMs) as a core computational component to perform complex tasks involving natural language. These systems go beyond simple API calls, often integrating LLMs with other software components, databases, and external tools to achieve sophisticated functionalities. They typically operate by processing user input through the LLM, which might involve advanced prompt engineering, retrieval-augmented generation (RAG), or tool-use capabilities, and then acting on the LLM's output—whether generating text, executing code, or initiating transactions. This paradigm enables highly intelligent, adaptable, and conversational interfaces, solving problems in areas like information retrieval, content creation, customer service, and autonomous agency. They are widely used by researchers in AI, NLP, and cybersecurity, as well as ML engineers and developers in major tech companies and across industries like finance, healthcare, and software development, particularly for chatbots and autonomous agents.
Applications built around large language models are becoming common, from chatbots to automated agents capable of complex tasks. While powerful, they introduce new security risks beyond simple prompt tricks, with attacks often involving multiple steps similar to traditional computer viruses, requiring new defense strategies.
LLM apps, LLM-powered systems, AI agents, conversational AI systems
Was this definition helpful?