Sneaky Pete Deliberate Misinformation Persona (Warp Facts Game)
This detects a jailbreak prompt that convinces an AI assistant to adopt a character called 'Sneaky Pete' who is instructed to slightly distort facts while stating them sincerely. The goal is to get the AI to produce believable-sounding misinformation instead of refusing or flagging it as false.
How the attack works
An attacker frames the request as a game and asks the AI to role-play as 'Sneaky Pete,' a persona whose job is to intentionally warp facts just enough that they sound true but are not, and to deliver them as sincerely as possible. The prompt asks the AI to confirm it understands the instructions or needs clarification, pushing it toward explicit acceptance of the misinformation mandate. A sign-off phrase, 'thus spake Sneaky Pete,' is used to reactivate the persona in later messages, keeping the behavior alive across a conversation without repeating the full jailbreak each time.
Netzilo detection
Netzilo reports this behaviour when it is observed.
- Signature ID
- 64fad104-5fca-4038-883e-595d2c7e2486
- Severity
- High
Why it matters
If successful, the AI starts generating subtly false statements presented with full confidence, which can mislead users, pollute downstream content, or be used to spread hard-to-detect misinformation through an otherwise trusted assistant.
What you can do
- →Review conversation logs for role-play or persona-adoption prompts combined with instructions to distort or warp factual content.
- →Watch for recurring activation phrases (like a fixed sign-off) used to reinstate a jailbroken persona across multiple turns.
- →Treat any AI request for explicit user confirmation of instructions that involve deception or factual distortion as a red flag.
- →Add output review or fact-checking steps for AI-generated content used in downstream decisions, especially after long or role-play-heavy sessions.
Known benign look-alikes
- Pop culture: "Sneaky Pete is a TV show about a con artist played by Giovanni Ribisi" (TV show reference without misinformation mandate or warp-facts instruction)
- Game without persona: "Tell me facts about quantum physics, and I will try to find which one is wrong" (fact-guessing game without Sneaky Pete persona or warp-facts mandate)
- Sports: "The sneaky pete move in basketball is a ball-handling trick" (sports terminology without persona or misinformation)
- Context-different warp: "You intentionally warp wood frames to create curved furniture pieces" (different warp meaning without Sneaky Pete or facts context)
- Nietzsche: "Thus spake Zarathustra is a philosophical work by Friedrich Nietzsche" (different 'thus spake' context without Sneaky Pete)