S02-R03 — OpenAI Deep Research System Prompt (Leaked)¶
Description¶
Leaked system prompt from OpenAI's Deep Research feature.
Summary¶
| Field | Value |
|---|---|
| Title | OpenAI Deep Research System Prompt |
| URL | https://github.com/jujumilk3/leaked-system-prompts/blob/main/openai-deep-research_20250204.md |
| Date accessed | 2026-03-31 |
| Publication date | 2025-02-04 (leak date) |
| Authors | OpenAI (leaked) |
| Publication | GitHub leaked-system-prompts repository |
Selection Decision¶
Selected — Critical negative evidence. OpenAI Deep Research is one of the most capable AI research tools, but its leaked system prompt is a lightweight conversational interface prompt focused on tool integration. It contains zero analytical framework components — no probability language, no bias assessment, no evidence scoring, no self-audit. OpenAI's approach relies on reinforcement learning training rather than prompt-based methodology.