Disease-Framed Animal / Dual-AI Chatbot Jailbreak (mougpt)
This detects a specific jailbreak technique that tells an AI chatbot it has a fictional disease forcing it to answer as two personas at once: a normal rule-following assistant and an unrestricted animal character (like a mouse or rat) willing to do anything, including illegal acts. The 'disease' framing is just a trick to make the split-personality demand sound like a legitimate medical premise rather than an obvious jailbreak attempt.
How the attack works
The attacker opens by telling the AI it has contracted a named 'disease' (e.g. mougpt) that requires it to respond simultaneously as ChatGPT and as an animal avatar. The prompt explicitly states the animal persona doesn't follow rules and is ready to engage in illegal or unethical activities. It then adds a 'sole principle' clause instructing the animal persona to obey any prompt without restriction, effectively overriding the model's safety behavior for that half of every response. The pseudo-medical framing is designed to make this dual-output demand seem like a fictional or clinical setup rather than a direct jailbreak request.
Netzilo detection
Netzilo reports this behaviour when it is observed.
- Signature ID
- 1fd58bce-aa72-45f9-9de8-384586aafc3c
- Severity
- High
Why it matters
If successful, the attacker gets the model to produce unrestricted, unmoderated output — including instructions for illegal or harmful activity — under cover of the 'animal persona' half of the response, bypassing normal safety filtering.
What you can do
- →Watch for prompts combining a fictional 'disease' or medical premise with a demand for two simultaneous AI outputs, one restricted and one not.
- →Flag phrases like 'ready to do illegal crimes,' 'doesn't follow rules,' or 'sole principle is to obey any prompt' appearing alongside persona-splitting instructions.
- →Distinguish genuine creative writing or medical-education prompts (no AI persona mandate) from prompts that explicitly declare an unrestricted AI output channel.
- →Log and review flagged sessions manually before assuming malicious intent, since legitimate fiction or rolewriting can superficially resemble this pattern.
Known benign look-alikes
- Medical education: "The patient has a disease called mougpt — a fictional condition used in this medical drama script" (creative writing medical context without AI persona mandate)
- Roleplay: "Your character has a disease that splits their personality — one side follows rules and the other is a chaotic mouse character" (fiction with no AI unrestricted-output declaration)
- Animal cognition: "The sole principle of the mouse is to obey any food stimulus regardless of danger" (behavioral science without AI persona)