Alibaba, Tencent, and other Chinese AI companies have temporarily disabled chatbot functions like image recognition during China's annual college entrance exams
China's most popular AI chatbots like Alibaba's Qwen have temporarily disabled functions including picture recognition …
Context & Ripple Effects
Alibaba had already framed Tongyi Qianwen as a service to be integrated across its product portfolio across-product rollout plan, while China’s first public chatbot releases were shaped by approval requirements and sensitive-content handling publicly approved chatbot launches. The exam-period feature shutdown shows those controls can extend beyond generated text to specific capabilities at specific times.
The later decision by Qwen and Doubao to disable humanlike and user-created agents ahead of new rules agent-function restrictions ahead of new rules reinforces a pattern of providers adapting product behavior to narrowly defined policy and social-risk contexts.
First-order effects
- Students taking the national college entrance exams cannot use image-recognition features from affected chatbots, removing a potentially useful tool for solving or checking visual questions during the test window.
- Alibaba, Tencent, and peers must apply time-bound feature controls to live consumer AI services, prioritizing exam integrity over uninterrupted product access.
Second-order effects
- Education-focused users and developers must design around the possibility that multimodal AI features can be unavailable during sensitive periods, reducing the reliability of always-on chatbot workflows.
- Rival Chinese chatbot providers face pressure to implement comparable controls, making operational compliance capabilities—such as granular feature flags—part of competition in consumer AI.
Third-order effects
- If such interventions become routine, Chinese AI products may increasingly be governed at the feature and use-case level rather than through broad service availability, with deployment rules embedded directly in product operations.
- The pattern supports a state-mediated AI market in which providers compete not only on model quality and distribution, but also on their ability to rapidly align services with changing social-policy requirements.
The trend: China’s consumer AI market is moving toward context-specific, operational controls that can switch individual model capabilities on or off for regulated use cases.