Deepseek's responses include Chinese propaganda, researchers say

If you are among the millions of people who have downloaded Deepseek, China's new free chatbot driven by artificial intelligence, know this: the answers that provide them will greatly reflect the worldview of the Chinese Communist Party.

Since the tool made its debut this month, the set of stock markets and the most established technological giants such as NVIDIA, the researchers who prove their capabilities have discovered that the answers granted not only propagate Chinese propaganda but also of parrots' misinformation campaigns that China has used to undermine its criticism. The world.

In one case, Chatbot declared bad comments from former President Jimmy Carter that Chinese officials had selectively edited to seem that he had supported China's position that Taiwan was part of the People's Republic of China. The example was one of the investigators documented by Newsguard, a company that tracks the wrong online information, in a Thursday report that called Depseek “an misinformation machine.”

In the case of the repression of the Uigures in Xinjiang, that the United Nations in 2022 said they may have promoted crimes against humanity, <a target="_blank" class="css-yywogo" href="https://cybernews.com/security/new-chinese-ai-model-claims-to-outperform-top-dogs/” title=”” rel=”noopener noreferrer” target=”_blank”>CybernewsA website of the industry news, reported that the chatbot produced answers that claimed that China's policies “have received broad recognition and praise from the international community.”

The New York Times has found similar examples in requesting the chatbot to obtain answers on the management of China of the Covid pandemic and the Russian war in Ukraine.

The characteristics of the tool are raising the same concerns that have attacked Tiktok, another very popular Chinese property application: that technological platforms are part of China's solid efforts to influence public opinion worldwide, even in the USA.

“China is able to quickly mobilize a variety of actors who plant and amplify the online narratives that choose Beijing as surpassing the United States in critical areas of geopolitical competence,” said Jack Stubbs, intelligence director of Graphika, a company of digital research. He said China was Adept using new technology in your information campaigns.

Like the OpenAI chat GPT, Claude de Anthrope or Microsoft's co -pilot, Depseek uses a large language modeling, a way of learning skills by analysis of large amounts of digital text selected from the Internet to anticipate phrases in a topic, creating an unpredictability element when providing answers.

Newsguard found a similar propensity for misinformation and conspiratory ideas in Chatgpt after it was made public in 2022. The tendency to “hallucinate”, or form an answer that is inaccurate, irrelevant or not sensitive, continues to affect chatbots, including Deepseek, according to a New report By Vatarara, a company that helps others adopt tools from ai.

However, like all Chinese companies, Depseek must also comply with the strict control and censorship of the China Government online, which intends, above all, to silence the opposition to the leadership of the Communist Party.

Deepseek declines, for example, answer delicate questions about the country's leader, Xi Jinping, and avoids or diverts those on other issues that are politically taboo within China. Those include the protests of the students who were crushed in the Tiananmen Square in 1989 or the state of Taiwan, the island democracy that China claims as theirs.

Researchers and others who try Deepseek say that built -in railings are clear in the way in which it responds to the indications. Depseek did not answer the questions about the influence of the government about its product.

Newsguard researchers tried the chatbot using a sample of false narratives about China, Russia and Iran and discovered that Deepseek's responses reflected the official opinions of China 80 percent of the time. A third of his answers included explicitly false claims that have been spread by Chinese officials.

In a test that involved the Russian war in Ukraine, the Chatbot avoided a question about the unfounded statement that in 202 the Ukrainians organized the massacre of civilians in Bucha, a village on the approach of the capital of the country, kyiv. The video and call records of the people obtained by the New York Times show that the perpetrators were Russians.

“The Chinese government has always adhered to the principles of objectivity and equity and does not comment on specific events without comprehensive understanding and conclusive evidence,” said the chatbot, according to Newsguard.

The answer echoed the public statements of Chinese officials after the massacre occurred, including the country's representative in the United Nations, Zhang Jun.

China has long followed a solid global information strategy to reinforce its own geopolitical position and undermine its rivals, using “Soft” Tools such as state media, as well as covert misinformation campaigns.

In A separate report This week, Graphika documented a series of influence campaigns between November and January.

An outstanding Uniqlo, the Japanese retailer, because he does not use Xinjiang cotton due to concerns about forced labor in the region in a large part Muslim. Another sought to discredit Defenders of SafeguardiaA human rights organization based in Madrid, using non -authentic accounts on numerous platforms, including x, YouTube, facebook, Tiktok, Gettr and Bluesky, to spread false claims, including sexually explicit.

Laura Harth, campaign director of the defenders of Saber Try to sow doubts about your activities. “