Key Takeaways
Zhipu AI has unveiled a new free browser assistant that competes with recent AI startups DeepSeek and Manus.
The new generation of Chinese AI platforms emphasizes efficiency and agentic capabilities. However, Zhipu and its peers are increasingly being targeted by U.S. sanctions.
Initially launched in October but opened to the public for free on Monday, March 31, AutoGLM leverages Zhipu’s GLM language model and is the company’s most ambitious consumer platform yet.
In line with the rising trend of AI computer use, AutoGLM is designed to operate on graphical user interfaces without the need for custom integrations.
In theory, that means the AI can “accomplish anything that humans can do on visual electronic devices,” Zhipu said in a statement .
The new tool is available for free via the Zhipu Qingyan browser extension and an Android mobile app.
In recent months, Chinese AI developers have made significant progress in efficiency.
With the launch of its R1 model in January, DeepSeek made waves for delivering cutting-edge capabilities at a much lower cost than leading Big Tech models.
Reflecting the heightened focus on efficiency, Zhipu’s creators highlighted techniques and design choices that made the most out of AutoGLM’s training hours, enabling it to complete tasks faster and with fewer resources.
In January, Zhipu was added to a list of restricted entities maintained by the Commerce Department’s Bureau of Industry and Security (BIS).
The designation prevents the startup from buying American technology, including Nvidia GPUs that are highly sought after by AI firms.
Other Chinese companies on the list include Huawei, SMIC and ZTE Corporation. In the past couple of years, AI startups like Zhipu, Megvii and SenseTime have also been added.
Although the U.S. has wielded the BIS blacklist in a bid to slow China’s AI advancement, there is mounting evidence that the strategy has backfired.
Far from preventing Chinese firms from developing powerful AI, U.S. sanctions helped catalyze the country’s recent AI boom by forcing developers to work around hardware limitations.