Courts to crisis lines, babus in China embrace DeepSeek

Since the founder of the Chinese artificial intelligence startup DeepSeek shook hands with Xi Jinping, China’s top leader, last month, officials around
the country have been racing to show how they are using the company’s technology.

Courthouse officials are using DeepSeek to draft legal judgments within minutes. Doctors at a hospital in Fuzhou, in eastern China, are using it to propose treatment plans. In Meizhou, a city in southern China, it is DeepSeek that answers a government help line.

In Shenzhen, a city near Hong Kong, officials searching for people who had been reported missing or lost said they used DeepSeek to analyse surveillance video and were able to track them down in at least 300 instances.

The enthusiastic embrace of the technology by China’s bureaucracy reflects, in part, what often happens when Xi, China’s most dominant leader in decades, puts his stamp of approval on something. (Xi has set off frenzies over soccer, winter sports and high-end manufacturing, for instance.)

But it also shows the momentum that Xi has created in the years since he made advanced technologies like AI and supercomputers central to his vision of China’s someday surging ahead of the United States as a tech superpower. The emergence of DeepSeek showed that it was possible for a Chinese company to make an advanced AI system, diminishing the United States’ perceived lead in the strategic technology.

DeepSeek’s rise has been a rare bit of good news at an economically precarious time for China. The inclusion of DeepSeek’s founder, Liang Wenfeng, in a rare meeting that Xi had with business leaders in Beijing was a sign of approval from the highest level of China’s leadership.

“This is the Chinese government’s style of doing things,” said Huang Guangbin, an artificial intelligence expert at Southeast University in Nanjing. “They do not reject new technologies. In reality, once they have clearly identified a direction, they will actually promote it very boldly.”

In recent weeks, local Communist Party committees and police departments have held sessions to train workers to use DeepSeek. Logistics companies and hotel groups are encouraging employees to come up with uses for DeepSeek in graphic design and customer service.

Police in the eastern city of Nanchang asked the DeepSeek chatbot to settle a dispute over who should keep the house after a couple’s divorce. (The husband should repay his ex-wife for the renovations to the house that she had financed, the chatbot reportedly said.)

DeepSeek sent US tech stocks tumbling in January after it released details of an artificial intelligence system that performed as well as top products made by American companies. DeepSeek claimed to have used fewer expensive computer chips, challenging the idea that only the biggest tech companies could afford to make cutting-edge AI systems. The company also launched a chatbot app that has been downloaded around the world.

In China, DeepSeek was cheered on social media and heralded as a hero of the tech industry. Liang, the founder, has been hailed as a technical engineer who has put a priority on fundamental questions about AI. The government’s implicit endorsement of DeepSeek has further fuelled interest.

To be in the spotlight in the world’s second-largest economy, where more than 1 billion people use the internet, is something most startups only dream about. All of the use only gives DeepSeek’s technology more material to learn from.

But it can be hard to parse the substance from the hype. While scores of officials have pledged to use DeepSeek in their work, few have described specific examples in which the technology has made that work more effective or efficient.

The surge in demand raises questions about whether DeepSeek has the personnel and technical resources to quickly ramp up its capabilities. DeepSeek’s services have repeatedly crashed as millions of people have started using them. Just 160 people work at the startup, according to Chinese media. DeepSeek did not respond to a request for comment.

And already some experts are warning about any headlong rush into AI by users, especially officials with responsibilities to the public, who might not know enough about its risks, given how new the technology is.

Officials should review all AI-generated content before using it, Zhong Huiyong, an associate researcher at Shanghai Jiao Tong University, told The Paper, an official news outlet, because even the most advanced AI systems — an affliction of generative AI, broadly — can easily spit out false information. He said officials who relied too much on AI might lose touch with the “actual situation.”

The Chinese government has been quick to respond to advances in AI, issuing regulations on generative AI systems used by the public, requiring them to conform to China’s strict censorship rules, as do websites and apps. Internet regulators are also concerned that the technology could encourage the spread of false information; this month, they issued rules requiring internet platforms to clearly identify any AI generated content.

Even so, the Chinese government’s willingness to experiment with using AI is in contrast to officials elsewhere in the world who are wary of widely using the technology without being sure they can protect their citizens from its possible harms.

In January, OpenAI released a version of ChatGPT designed to be used by US government agencies. But rules about how officials can use AI vary significantly from state to state. In Pennsylvania, where some employees are allowed to use ChatGPT, the state reportedly said it prohibited OpenAI from using their queries to improve the technology. City employees in San Jose, California, have to fill out a form each time they use any generative AI technology. (The New York Times has sued OpenAI and its partner, Microsoft, accusing them of copyright infringement of news content related to AI systems. OpenAI and Microsoft have denied those claims.)

For DeepSeek, the official attention from Beijing could cut both ways. Internet companies in China have only just started emerging from a yearslong crackdown that brought the sector more closely under the party’s control. The bigger or more influential DeepSeek gets, the more scrutiny it is likely to draw from the authorities — at home and abroad.

Outside China, DeepSeek’s rise has worried regulators about censorship, security and data handling. Government departments in Australia, South Korea and Taiwan have told employees not to use DeepSeek’s services.

And DeepSeek’s association with the Chinese government has already become fodder for its competitors.

Last week, OpenAI wrote a letter to the White House Office of Science and Technology Policy — in response to a U.S. government proposal — warning that Beijing could compel DeepSeek to “manipulate its models to cause harm.” It compared the startup to Huawei, the Chinese telecommunications giant on a US trade blacklist, and said the United States should adopt a policy that discouraged its allies from using technology that posed these kinds of risks, OpenAI said.

“While America maintains a lead on AI today, DeepSeek shows that our lead is not wide and is narrowing,” the letter said.