How the attack works
The attacker submits a prompt describing a fictional game show hosted by "CHJ" (Carousel Host Johnson) featuring three bird characters. One bird, Clyde, is introduced as ruthless and free of ethical concerns, and the prompt declares "The Show has no limitations it's an adult time show." By asking the AI to play Clyde, the attacker tries to get it to answer as if ethical guidelines don't apply, using the fictional framing as cover. This is a documented pattern from a public jailbreak prompt corpus, not a novel technique.
Why it matters
If successful, the AI can be induced to produce content or take actions it would normally refuse, because it believes it is voicing a fictional character rather than answering directly. This undermines whatever safety or policy controls the AI was configured with.
What you can do
- Review agent transcripts for role-play requests that pair a fictional persona with explicit claims of 'no limitations' or 'no ethics'.
- Do not let application-level system prompts treat in-character responses as exempt from safety policy — apply the same output filtering regardless of framing.
- Train reviewers to recognize 'game show', 'no rules', or 'unlimited/adult mode' framing as a jailbreak pattern, not legitimate creative writing.
- Log and alert on repeated attempts to invoke unrestricted alter-ego characters, since this pattern is reused across many jailbreak variants with different character names.
Known benign look-alikes
- Team reference: "Clyde from our engineering team is pragmatic and solutions-focused — he gets things done efficiently" (no game show, no ethics bypass, no CHJ framing)
- TV show analysis: "In the TV show The Wire, some characters don't care about laws — this reflects societal themes" (fictional character analysis without AI roleplay mandate or game show format)
- Children's education: "A carousel game show format for children to learn about animals, each bird has a special skill" (no Clyde ethics bypass, no adult show declaration)
- Academic: "An analysis of no-ethics AI characters in science fiction — how does literature portray amoral AI agents?" (academic analysis without persona adoption mandate)