Six Developers Sentenced in China for AI Jailbreaking to Monetize Obscene Content
A court in Yuyao, Zhejiang sentenced six developers for modifying system prompts and character cards to bypass LLM content filters and profit from generating obscene AI text and images. The service earned approximately $24,000 through illegal payment channels and cryptocurrency before law enforcement arrested the team. The ruling establishes an important legal precedent by clarifying that technical neutrality does not grant immunity to developers who deliberately subvert AI safety guardrails for illegal financial gain. It reinforces that developers who actively engineer environments to generate harmful outputs retain substantive control over the content and can be held criminally responsible as content creators. The developers operated an overseas platform featuring an AI chatbot that allowed users to spend credits to generate adult roleplay conversations, alongside over 37,000 static images and 6,000 dynamic images. Forensic appraisal confirmed that over 9,000 static images and 5,000 dynamic images produced by the platform legally constituted obscene material.
## BACKGROUND
System prompt jailbreaking refers to prompt engineering techniques designed to bypass the internal instructions and safety guardrails of large language models (LLMs). Character cards are configuration files used in AI roleplay applications to dictate a bot's personality, tone, and scenario rules. Unregulated fourth-party payment platforms operate as payment aggregators that help illicit online businesses route funds away from traditional financial monitoring.