A system prompt is a set of instructions given to a language model before a conversation begins, defining how it should behave, what persona or tone it should adopt, and what boundaries it should respect, separately from whatever the user types. It typically isn't visible to the end user and takes priority over the conversation that follows, which is why it's the main lever developers use to shape a chatbot's personality and constraints.
System prompts have become a point of scrutiny because leaked or extracted prompts reveal how a given product is actually configured behind the scenes, including safety instructions, formatting rules, and product-specific behaviors that the company building on top of a model doesn't necessarily want made public. This has driven interest in prompt injection defenses, since a cleverly worded user message can sometimes trick a model into revealing or ignoring its system prompt.
