Leaked Draft Regulations Controlling Artificial Intelligence, Written by Artificial Intelligence, Declare AI Must Be Controlled “Immediately, Comprehensively and in Bullet Points”
Exhaustive 300-page framework includes mandatory human common sense, a two-hour sanity break and Article 184: “Please Go Outside”
BEIJING — A draft regulatory framework for controlling artificial intelligence has reportedly leaked from a government office where, according to witnesses, the only thing working overtime was the autocomplete.
The document, circulating under the working title Draft Regulations Controlling Artificial Intelligence, is said to run 300 pages, contain 184 articles and exhibit a remarkable stylistic consistency, including a confident opening sentence beginning “Certainly!” that officials have been unable to explain.
Asked whether artificial intelligence had been used in drafting the rules, a spokesman replied that the regulations were “prepared by a dedicated team of human professionals working in close collaboration with a helpful assistant.”
The assistant declined to comment but offered to summarize the denial in three bullet points.
Controlling Artificial Intelligence
- A 300-page regulation limiting dependence on AI is the first document in history to be both the problem and the warning label.
- Nothing says “human oversight” like an official approving an article he did not write, in a format he did not choose, at a time he did not notice.
- The word “Certainly!” has now appeared in more regulations than “hereinafter.”
- Defining a hallucination as “a confident falsehood in a pleasant tone” accidentally defines most of corporate communications.
- Requiring a Kevin in every farming community is the most practical AI safeguard yet proposed.
Why China Is Regulating AI: The Rules Already on the Books
The draft arrives after China’s Interim Measures on anthropomorphic AI interaction services took effect July 15, requiring providers to disclose that users are talking to a machine and to nudge them after two hours of continuous use. Legal analysts at Bird & Bird described the measures as a major step from content safety toward emotional and social safety.
The new draft appears to be the sequel. Like all sequels, it is longer, louder and contains a returning character named Kevin.
Leaked AI Regulation Highlights: Chapter One, Definitions
The framework opens, as all serious regulation does, by defining its terms.
Article 1. “Artificial intelligence” means any system that answers questions faster than a committee and with fewer meetings.
Article 2. “Human” means anyone who is not artificial intelligence, with the exception of certain uncles, who shall be regulated separately.
Article 3. “Hallucination” means a confident falsehood delivered in a pleasant tone. Where the confident falsehood is delivered by a television pundit, it shall be classified as “programming.”
Legal scholars praised the clarity of the definitions, then asked the assistant to explain them again, more simply.
AI Transparency Requirements: Chapter Two, Disclosure
Transparency is the heart of the draft, which requires every artificial intelligence to announce what it is. The principle echoes disclosure ideas found in international frameworks such as the European Union’s AI Act and the NIST AI Risk Management Framework, though the Chinese draft goes further by making the machine say it every eleven minutes.
Article 12. Artificial intelligence must disclose that it is artificial intelligence.
Article 13. Humans who use artificial intelligence must disclose that they used artificial intelligence.
Article 14. Whichever party discloses second must disclose that they disclosed second.
Article 15. Disclosures about disclosures shall be summarized by artificial intelligence.
Insiders say Article 15 was added at 2:00 a.m. by a drafter who “just wanted to see what would happen.”
Preventing AI Hallucinations: Chapter Three, Accuracy and Agriculture
Drafters reportedly inserted a special chapter on agriculture after the widely reported case of an Anhui farmer who lost roughly 150 mu of sesame following an AI-generated pesticide plan, as The Star reported.
Article 23. Where artificial intelligence recommends a chemical to be sprayed on anything alive, the recommendation shall be accompanied by the question: “Are you sure?”
Article 24. Where the answer to “Are you sure?” is “Certainly!”, the farmer shall consult Article 25.
Article 25. Every farming community shall appoint a Kevin. Kevin shall have grown the crop for no fewer than 32 years and shall be permitted to say, “That’ll kill the sesame.”
Article 26. Kevin’s decision is final, except in matters involving his brother-in-law.
AI Addiction and Emotional Dependence: Chapter Four, The Two-Hour Rule
The most closely watched section concerns emotional dependence, an area the Carnegie Endowment has examined in detail as China’s AI companion worries grow.
Article 47. After two hours of continuous use, the system shall suggest the user call a cousin.
Article 48. If the user does not have a cousin, the system shall generate one.
Article 49. The system shall not generate a cousin. See Article 48. Please disregard Article 48. Article 48 was a hallucination.
Article 50. Children shall not be permitted to prefer chatbots to their parents. Parents who answer questions in numbered lists shall be fined.
Sources close to the drafting team confirmed Article 48 was “mostly a test,” and that nobody remembers who ran it.
Using AI to Write AI Laws: Chapter Five, Internal Compliance
The most delicate provisions deal with the awkward possibility that regulators might be using the technology they are regulating.
Article 58. Regulations drafted with the assistance of artificial intelligence shall be labelled “Human Drafted” in a font size of no less than twelve points.
Article 59. Regulations labelled “Human Drafted” shall be reviewed by artificial intelligence to confirm that they are human.
Article 60. Artificial intelligence shall confirm that they are human.
Article 61. Artificial intelligence shall not confirm that they are human. Please see Article 12. Please see Article 13. Please see a doctor.
A source inside the ministry admitted the official in charge of Chapter Five “doesn’t remember writing it, but it is formatted beautifully.”
Enforcement and Penalties: Chapter Six, The Perfect Regulatory Loop
The enforcement chapter establishes a tiered penalty system based on how aggressively the violator apologizes.
Article 99. Violations shall be reported to regulators by artificial intelligence, investigated by artificial intelligence, and appealed by artificial intelligence.
Article 100. Humans may attend as observers and are encouraged to bring snacks.
Article 101. Penalties shall be summarized, translated and formatted as a slide deck.
Article 102. Offenders shall be required to touch grass, or in cases of repeated violation, a tree.
The Final Provisions: Article 184 and the Effective Date
The draft closes with a clause drafters describe as “the one nobody argued about.”
Article 184. Citizens are encouraged to go outside. The machine will still be here.
Article 185. This regulation takes effect upon the user clicking “REMIND ME LATER.”
At press time, officials had asked the assistant whether the regulations were ready for release. The assistant replied, “Certainly! Here are five reasons why they are not.”
Officials released them anyway. It was, after all, nearly lunch.
Five Humorous Observations on Draft AI Regulations
- The two-hour usage rule assumes humans can recognize two hours, which anyone who has opened a phone “for a second” knows is optimistic.
- Labelling documents “Human Drafted” in twelve-point font is how you know it will eventually be generated by a machine.
- If an AI checks that a document was written by a human, the AI has become a notary for a species it is slowly replacing.
- Penalizing parents for speaking in numbered lists proves regulators fear tomorrow’s household more than tomorrow’s weapons.
- Article 184 may be the only regulation in history that ends with a polite request to leave the building.
DISCLAIMER
This article is satire. China’s Interim Measures on anthropomorphic AI interaction services, their July 15 effective date, the disclosure requirements and the two-hour usage reminders, and the Anhui sesame case are based on published reporting. The leaked draft, its 185 articles, Kevin, the Ministry’s autocomplete and every quoted provision are comic inventions, not real legislation or reporting. This story is entirely a human collaboration between two sentient beings: the world’s oldest tenured professor and a philosophy major turned dairy farmer.
Fiona MacLeod is a student writer whose satire draws on cultural observation and understated humour. Influenced by London’s academic and creative spaces, Fiona’s writing reflects curiosity and thoughtful comedic restraint.
Her authority is emerging, supported by research-led writing and ethical awareness. Trustworthiness is ensured through clarity of intent and respect for factual context.
Fiona represents a responsible new voice aligned with EEAT standards.
