Image by Wei Yu, from Unsplash

AI Tools Disabled Across China To Keep Students From Cheating During Exams

• Written by Kiara Fabbri Former Tech News Writer
• Fact-Checked by Sarah Frazier Former Content Manager

The Chinese AI tools, Alibaba’s Qwen and Tencent’s Yuanbao, disabled their photo recognition features as a temporary measure to prevent students from cheating during the national gaokao college entrance exams.

In a rush? Here are the quick facts:

• China’s gaokao exams run from June 7 to 10 this year.
• AI tools like Qwen and Yuanbao disabled photo features during exams.
• Chatbots blocked image recognition to prevent student cheating.

Some of China’s most popular AI tools, including Alibaba’s Qwen and Tencent’s Yuanbao, have temporarily shut down key features, as first reported by Bloomberg . These features include photo recognition to stop students from cheating during the country’s high-stakes college entrance exams, known as the gaokao.

The gaokao, which takes place from June 7 to June 10, stands as the most crucial examination for Chinese students throughout their academic journey. The annual test attracts 13.4 million students, noted Bloomberg.

The exam determines whether a teenager will secure admission to a prestigious university or must spend another year in high school. The gaokao represents the sole opportunity for numerous students from small towns and low-income families to move forward in their lives, says Bloomberg.

The photo-scanning functions of Tencent’s Yuanbao and Moonshot’s Kimi became unavailable to users during examination periods. Bloomberg reports that when users tried to access them, the chatbots replied: “To ensure the fairness of the college entrance examinations, this function cannot be used during the test period.”

The photo-scanning feature of Alibaba’s Qwen remained accessible to users, yet the system refused to process test images during examination hours. ByteDance’s chatbot Doubao responded to a photo of an exam by saying the image was “not in compliance with rules,” as reported by Bloomberg.

The government has implemented more stringent regulations to monitor the exams. Students are prohibited from bringing phones and devices into exam halls where the gaokao takes place, says Bloomberg. The recent rise of AI tools, which solve complex problems from images, has created fresh concerns about digital cheating.

Bloomberg reports that the Chinese education ministry issued a warning to schools in May about preventing students from using AI-generated content in their homework assignments and tests. Bloomberg notes authorities simultaneously promoted the development of AI competencies among young students.

The Chinese examination process operates differently from the American system, which uses grades along with essays and activities for college admission decisions. The Chinese educational system places such importance on this single test that it determines students’ future paths, which makes fairness and AI control essential national priorities.

Photo by Shawn Tung on Unsplash

Chinese Social Media Platform RedNote Releases Open-Source AI Model

• Written by Andrea Miliani Former Tech News Expert
• Fact-Checked by Sarah Frazier Former Content Manager

The Chinese social media platform Xiaohongshu, known internationally as RedNote, released its first open-source large language model (LLM), called dots.llm1, on Friday. The company claimed the model matches the performance of the leading AI models in the tech industry.

In a rush? Here are the quick facts:

• RedNote launched a new open-source AI model called dots.llm1.
• The company claims its MoE system outperforms DeepSeek-V3 and Qwen2.5.
• The new model is open-source and available on Hugging Face.

According to the South China Morning Post , the AI model released by RedNote is a mixture-of-experts ( MoE ) system—an architecture that includes multiple specialized sub-networks, or “experts,” designed for greater efficiency and to rival the most advanced AI models currently on the market.

RedNote explained that its AI research team developed dots.llm1 in collaboration with its in-house Humane Intelligence Lab. The company has been recruiting experts with strong backgrounds in the humanities to make sure its technology can better simulate and understand human behavior.

The company claims that dots.llm1 outperforms regional competitors, including Alibaba’s Qwen2.5, and DeepSeek-V3. RedNote also developed Diandian, an AI research assistant, which includes a “deep research” function.

According to Reuters , unlike most American companies—such as Google and OpenAI—, Chinese firms are relying on open-source principles to showcase their technological capabilities and expand their global influence.

Users can download the model via Hugging Face and explore its technical specifications.

“The dots.llm1 model is a large-scale MoE model that activates 14B parameters out of a total of 142B parameters, delivering performance on par with state-of-the-art models,” states the company’s page on Hugging Face. “Leveraging our meticulously crafted and efficient data processing pipeline, dots.llm1 achieves performance comparable to Qwen2.5-72B after pretrained on 11.2T high-quality tokens without synthetic data.”

RedNote currently has around 300 million monthly active users and recently reached a $26 billion valuation. The company expects to launch its IPO later this year.

In January, RedNote gained significant popularity among American users as an alternative to TikTok amid the potential U.S. ban on the video-sharing platform.