During this year’s university entrance exams in China, leading tech companies unexpectedly disabled key features of their generative AI. More than thirteen million applicants participating in the four-day test marathon were unable to use modern digital tools, which some participants might have relied on.
Screenshots appeared on online platforms where users tried to upload tasks to Doubao from ByteDance, but received the response: “During the university entrance exams, according to relevant requirements, the question-answering service will be suspended.” A similar reaction awaited those who turned to DeepSeek — here, the AI reported the unavailability of the service during specified hours to ensure the fairness of the exams. Tencent with Yuanbao, Alibaba with Qwen, and Kimi from Moonshot also temporarily disabled image recognition during exam hours.
This massive “freeze” of functions did not go unnoticed among students, who actively discussed the situation on social media. Some joked about the inconveniences, highlighting that they would now have to look for alternatives or return to other services. Meanwhile, companies refrain from official comments on the reasons and duration of such restrictions.
The disabling of generative AI capabilities is accompanied by other measures to combat potential violations — from using AI-enhanced video surveillance systems to detect suspicious behavior to enhanced entry checks, biometric identification, and radio signal blocking. In several cities, even mass events were postponed, office hours were changed, and separate transport corridors were organized to ensure calm and timely arrival of participants for the exams.
