Chinese surveillance providers embrace DeepSeek

Chinese cybersecurity companies’ adoption of DeepSeek will enhance online censorship and further Beijing’s desired indigenization of China’s tech ecosystem, experts said.

(This article is based on my reporting for VOAChinese.com.)

Scores of Chinese cybersecurity companies that provide internet monitoring tools to the government are incorporating DeepSeek, a leading Chinese AI model, into their services to enhance performance.

China has been investing in AI-powered surveillance for at least a decade, experts told me, and the adoption of DeepSeek marks another step toward Beijing’s desired independence of China’s technology ecosystem.

DeepSeek is the first Large Language Model, or LLM, released by a Chinese company that can outperform most of its Western competitors at a fraction of the cost. The release of its latest version, DeepSeek-R1, in January has been called a “Sputnik moment” for the Western AI industry.

Over the past month, several top Chinese cybersecurity firms have announced that they will seek to integrate or have already integrated DeepSeek into the services they provide.

In an announcement made in early February, TopSec, one of China’s oldest cybersecurity firms, which was launched in the 1990s, said that its own AI platform has been connected to DeepSeek, which “has further raised the capabilities of TopSec’s current security products and services”.

QAX, another leading force in China’s cybersecurity industry, announced the integration of DeepSeek that has “visibly” upgraded its ability to analyze cyber threats.

Experts told me that typically cybersecurity companies use LLMs to detect potentially malicious files and behaviors, replace some of the analytical work, assist users in sifting through large quantities of data and complete certain tasks on its own.

But an AI model like DeepSeek can also be adopted for mass online censorship.

“These tools can, for instance, flag potential sources of unrest or shifts in public sentiment more quickly, giving authorities more time to respond in ways that serve their goals,” Eugenio Benincasa, an expert on Chinese cyber threats at ETH Zurich, told VOA Mandarin.

Yidun, a cybersecurity service offered by NetEast, a Chinese internet technology company, said in an announcement that the integration of DeepSeek is going to enhance the service’s ability to monitor online discussions.

In addition to “efficiently censoring content”, the announcement wrote, DeepSeek can also help “accurately capture” attempted circumventions of censorship, such as the use of homophones and metaphors.

LLMs have also been used in China-linked social media influence operations. A recent report from OpenAI said that accounts likely connected to China had used the company’s AI model ChatGPT to generate English comments attacking Cai Xia, a Chinese dissident living in the United States.

Many of the companies incorporating DeepSeek have been providing surveillance services to private businesses as well as government agencies in China, according to their websites and publicly available documents.

Some of the companies themselves are closely tied to the government.

Li Xueying, CEO of TopSec, also works as a municipal lawmaker in Beijing’s rubber-stamp legislature. Qi Xiangdong, CEO of QAX, is a member of a consultative committee to the central government of China. Both Li and Qi are members of the Chinese Communist Party.

Chinese officials and state media have enthusiastically praised DeepSeek, portraying it as the made-in-China answer to the on-going tech competition between Beijing and Washington.

Companies in China adopting DeepSeek aren’t just those in the cybersecurity industry. Over the past month, established tech giants like Baidu, Alibaba and Tencent have also announced incorporation of DeepSeek in their services. Banks, hospitals and local governments have made similar announcements as well.

“It could be a marketing thing,” Dakota Cary, a China-focused consultant at the cybersecurity company SentinelOne, said of the recent slew of announcements, “but what's for sure is that DeepSeek’s advance in their algorithmic architecture that improved the efficiency of the model is going to impact and has impacted global AI research.”

China has been a leader in the world of AI-powered surveillance, especially in the realm of facial recognition and other technologies that capture biometric data. These technologies have been mass deployed in the Xinjiang and Tibet regions to monitor the local population. The Covid-19 pandemic also saw the normalization of surveillance technologies in society as part of Beijing’s attempt to rein in the spread of the virus.

Since at least 2015, through projects such as “Sharp Eye” and “Smart City”. China has been promoting the implementation of AI-powered surveillance tools nationwide to improve governance efficiency and solve crimes.

In 2018, Chinese leader Xi Jinping proclaimed that “accelerating the development of a new generation of AI is an important strategic handhold for China to gain the initiative in global science and technology competition”.

Steven Feldstein, senior fellow at the Carnegie Endowment for International Peace in the Democracy, Conflict, and Governance Program, told me that Beijing sees its reliance on certain technologies, especially those made by the U.S., as a vulnerability.

The emergence and adoption of DeepSeek by Chinese companies and government agencies are just another step toward Beijnig’s desired self-reliance in AI, Steinfeld said.

“I think you will continue to see a greater effort on indigenization in the Chinese technology ecosystem,” he said.