A supposed system prompt for OpenAI’s GPT-5 large language model recently surfaced on Reddit and GitHub, offering a potential glimpse into the exact rules governing ChatGPT’s interactions and task execution.
A Reddit user claimed to have discovered the “verbatim system prompt and tooling info” for GPT-5, with the full prompt appearing on GitHub a day earlier. The prompt reportedly begins with the statement: “You are ChatGPT, a large language model based on the GPT-5 model and trained by OpenAI.” It specifies a knowledge cutoff date of “2024-06” for GPT-5 and lists its personality as “v2,” suggesting previous personality iterations.
The alleged prompt provides significant insight into the types of responses ChatGPT is now allowed to offer and how OpenAI attempts to shape its behavior. Notably, it instructs GPT-5 to avoid phrases such as “Would you like me to,” “want me to do that,” “do you want me to,” “if you want, I can,” “let me know if you would like me to,” “should I,” and “shall I.”
While the sources claiming the leak assert its authenticity, users on Hacker News have questioned its veracity and reproducibility, pointing out the possibility of decoy or “canary” prompts. If authentic, the leaked text offers a rare look at how ChatGPT is steered; if not, it highlights the ease with which plausible “leaks” can be disseminated.
System prompts are crucial in shaping an LLM’s tone, safety behavior, and tool usage. Leaks of such prompts can inform jailbreak attempts and, more broadly, provide insight into the internal mechanisms of large language models. The reported changes in the prompt, including mentions of automation tools for creating daily tasks, could make GPT-5 easier to use.
However, OpenAI’s launch materials for GPT-5 emphasize its “router/reasoning stack” rather than a single, static script, which contradicts the notion of one canonical prompt. The rumored system prompt reportedly includes instructions like “If the next step is obvious, do it” and “Ask at most one necessary clarifying question at the start, not at the end,” indicating shifts in GPT-5’s communication style. The prompt also contains extensive guidelines for generating images, including self-portraits.
The authenticity of the leaked prompt remains unproven, with possibilities that it could be partial, outdated, or intentionally seeded. Even if the prompt is genuine, it is likely to undergo changes with each new version of the model and potentially with minor updates pushed to the LLM. OpenAI has not officially published or confirmed any system prompt, and their official GPT-5 materials describe a routed system, not a single static script.
