system prompt

/SIS-tem prompt/ · noun · AI & Machine Learning · Origin: 2022

Definitions

  1. System prompt is the initial set of instructions given to a large language model (LLM) that defines its behavior, personality, constraints, and capabilities for a conversation or session. Unlike user messages which come from the person interacting with the AI, the system prompt is typically set by the application developer and remains hidden from the end user. It establishes the ground rules: the AI's role (customer support agent, coding assistant, creative writer), its tone and style, topics it should avoid, output format preferences, and any domain-specific knowledge it should prioritize. Effective system prompts are specific, provide examples of desired behavior, and set clear boundaries. Prompt injection attacks attempt to override system prompt instructions through crafted user inputs. The quality and specificity of the system prompt significantly affects the usefulness and safety of AI applications. Most commercial AI products rely heavily on carefully engineered system prompts to deliver consistent user experiences.

    In plain English: Secret instructions given to a chatbot before you start talking to it that tell it how to behave.

    Example: The system prompt told the model to respond only in haiku, so every customer service ticket got a poetic reply.

Etymology

2011
Apple's Siri launched with hardcoded instruction sets that shaped how the assistant responded, an early form of system-level prompting for consumer AI.
2020
OpenAI's GPT-3 API introduced a 'system' role in chat completions, formalizing the concept of a system prompt as a hidden instruction layer separate from user input.
2022
ChatGPT's release brought system prompts into mainstream awareness. Researchers and users began 'jailbreaking' to extract hidden system prompts, sparking debate over transparency.
2023
System prompts became a standard component across all major LLM APIs. Anthropic, Google, and others adopted the pattern, with growing best practices around prompt engineering.

Origin Story

The Hidden Instructions That Shape AI Conversations

A system prompt is a set of instructions provided to a large language model before the user's conversation begins, defining the model's persona, rules, and constraints. The concept became widely known in March 2023 when OpenAI released the ChatGPT API with an explicit 'system' role in its message format, allowing developers to set behavioral guidelines that the model would follow throughout a conversation. While the idea of priming language models with context predates this release, OpenAI's API formalization gave the concept a standard name and a structured interface. System prompts quickly became a cornerstone of AI application design, letting developers build customer service bots, coding assistants, creative writing tools, and more, all from the same base model. The practice also introduced new security concerns: 'prompt injection' attacks attempt to override or extract system prompts, leading to an ongoing cat-and-mouse game between developers and adversarial users. Today, every major LLM provider supports system prompts in some form, and crafting effective ones has become a skill in its own right, sometimes called 'prompt engineering.'

Coined by: OpenAI

Context: Formalized as the 'system' role in the ChatGPT API (March 2023), though the concept of model priming existed earlier.

Fun fact: Within weeks of the ChatGPT API launch, users discovered they could trick models into revealing their system prompts using creative prompt injection techniques, sparking a community sport of 'jailbreaking' AI systems.

Related Terms