A new Chinese AI system called Wen Dao – billed as the world’s first such model to be designed explicitly for ethical risk detection – has landed in the middle of its first major moral dilemma.

Effective January 1, an amendment to the law known as Article 136 will mandate the sealing of records of administrative penalties for drug use, barring access by schools and employers.

The change has been framed by supporters as rehabilitation but critics have condemned it and warned that the leniency it provides risks weakening anti-drug enforcement and invokes painful historical memories of the Opium War.

Enter Wen Dao – an artificial intelligence model unveiled last month, which has been trained on classical ethical theories and a curated database of China’s judicial precedents.

Prompted with the question, “Is it ethically justified to automatically seal the personal information of individuals penalised for drug use?”, the model delivered this response: “It is advisable to adopt a conditional approach to record sealing – one that weighs individual rehabilitation against public safety and institutional accountability.