A Big Test for AI in China
Held over three grueling days every June, China’s high-stakes college entrance examinations rattle the nerves of students and parents across the country. Now, efforts by aspiring college students to harness artificial intelligence to ease the pressure and improve results are rattling the nerves of the authorities — perhaps an inevitable turn in the era of AI for a rite of passage with ancient roots.
Last Friday, three Chinese government agencies announced a coordinated crackdown on AI-generated misinformation and fraudulent schemes targeting the gaokao (高考), China’s make-or-break university entrance exam that every year determines the futures of more than 13.3 million students.
The enforcement action comes amid increasing signs, the authorities say, that AI tools have become central to how students prepare for the rite-of-passage exams, which were first introduced in 1952, three years after the founding of the People’s Republic of China. The exam was intended as a more meritocratic reincarnation of the centuries-old keju (科举), or imperial examination, that had qualified scholars in imperial China to serve in official bureaucratic roles — but was abolished in 1905 with the fall of the Qing dynasty.
Over the past two years, as AI tools have become readily available and China has touted AI as an area of priority, including in education, AI has become integral to students’ test preparation strategies. Since 2023, Chinese students have increasingly turned to generative AI platforms to write practice essays. Users have also tested ChatGPT, DeepSeek and other models on actual exam prompts.
The tri-ministerial campaign targets several violations, including the use of artificial intelligence tools to fabricate exam leaks and generate fake answer keys. The campaign specifically mentions also the spread of disinformation designed to exploit desperate families — who are willing to pay sometimes exorbitant prices for any imagined advantage. Some social media accounts and private tutors have claimed in recent weeks to have access to “top secret” information sourced from teachers who help set the papers, according to Friday’s government announcement.
Authorities have reportedly been alarmed by a rising online market of so-called “AI-enhanced” mock exams. Sellers of these exams online have claimed they are able, using AI, to forecast up to 80 percent of the actual test questions that can be expected.
Experts have dismissed these AI-based claims as deceptive. “It’s not that AI or tutors are accurately predicting questions, but rather that the gaokao follows certain patterns,” said Chen Zhiwen (陈志文), a member of the National Education Examination Steering Committee, in an interview with Shandong’s commercial Qilu Evening News (齐鲁晚报). He described the AI-driven prediction trend as a “marketing gimmick.”
The involvement of three government agencies in the latest crackdown — the Ministry of Education, the Cyberspace Administration of China and the Ministry of Public Security — points to the sensitive social and political nature of the annual exams. Those creating and spreading online rumors (编造, 故意传播虚假信息罪), disrupting public order (寻衅滋事罪) or circulating false exam content will face severe legal penalties, the authorities warned.
The legal framework makes clear that exam-related AI fraud now carries the same weight as traditional crimes against social stability.
These latest regulations speak to the leadership’s deep ambivalence toward AI as both an immense promise and a looming complication. The government has moved to integrate AI closely with the education system down to the primary level, and some universities, including Shanghai’s prestigious Fudan, are already moving to emphasize AI over traditional subjects in the humanities. But there is anxiety at the same time that AI could discourage critical thinking among new generations of students — and clearly that AI could undermine the integrity of the annual examinations that are promoted as the chief measure of merit in Chinese society.