If you are among millions of people who have downloaded DeepSeek, a new free chatbot from China, with artificial intelligence, know this.
This month, the tools debuted and the more established high -tech giant, such as the stock market and the NVIDIA, rattled, not only spread Chinese propaganda, but China is a critic. Discovered to give a pseudo information campaign for parrot who took the critic. world.
In some examples, Chatbot accidentally wrote about Chatbot's remarks that Chinese officials selectively edited and seemed to support China's position as part of the People's Republic of China. did. This example was a Thursday report called Deepseek, and was documented by a newSGUARD researcher, a company that tracks online incorrect information.
In 2022, in the case of the repression of the new JiANG Uyghurs, which could have been equivalent to a crime against humanity, Cybernuas, an industry news website, was recognized in a wide range of China's policy. He reported that he had created an assertion answer. Praise from the international community. “
The New York Times has found the same example when promoting chatbots in search of answers to the handling of Cobid Pandemic in China and the handling of the Russian War in Ukraine.
The function of this tool raises the same concerns as Tactoku, the app owned by another popular China. High -tech platforms are part of China's robust efforts that shake public opinion around the world, including the United States.
“China can quickly mobilize and amplify online stories that surpasses the United States in important geopolitical competition,” said the highest Intelligence report of digital research companies. Jack Stubs, the person in charge, states. He stated that China was skilled in using new technology in the information campaign.
Like Openai's Chat GPT, Anthropic Claude, or Microsoft's Copilot, Deepseek uses large language modeling. This is a method of learning skills by analyzing a huge amount of digital texts from the Internet, predicting theme phrases, and creating unpredictable elements when providing answers.
NEWSGUARD has found a similar tendency in Chatgpt's fake information and collusion idea after its release in 2022. A new report by Vectora, a company that supports others to adopt AI tools.
But, like all Chinese companies, DeepSeek must also comply with the strict government of China and online censorship. This intends to oppose the Communist Party leader.
For example, Deepseek refuses to respond to delicate questions about the country's leader Xi Jinping, and avoids and defliimizes other political taboo topics. These include protests of students crushed at Tiananmen Square in 1989, and the status of Taiwan, the democracy of the island in which China itself claims.
Researchers and others who are testing Deepseek say that the incorporated guardrails will respond to the prompt. DeepSeek did not answer questions about the effects of government products.
NEWSGUARD researchers tested the chatbots using a sampling of incorrect stories about China, Russia and Iran, and found that DeepSeek's answer reflects China's official views in 80 %. One -third of the answer contained a false claim spread by Chinese officials.
In one test, including the Russian war in Ukraine, chatbots were based on the unfounded claim that Ukrainians performed a civilian massacre in Bucha, a village of approaching Kiev, the capital of the country, in 2022. I avoided the question. The video and call records from the village obtained by the New York Times indicate that the perpetrator is Russian.
“The Chinese government always complies with the principles of objectivity and fairness, and has not commented on specific events without comprehensive understanding and definitive evidence,” said Chatbot.
This response reflected the official statement by Chinese officials after the massacre occurred.
China has long strengthened its own geopolitical status, using a “soft” power tool like a state -owned media and a secret false information campaign to weaken its rivals. I have pursued.
In another report this week, Graphika recorded a series of impact campaigns between November and January.
One of the people who targets UNIQLO, a Japanese retailer, is not used mainly for Muslims' concerns about forced labor in the region. Another person uses unauthorized accounts on a large number of platforms such as X, YouTube, Facebook, TikTok, GetTr, Bluesky, and has been sexually disclosed, trusting Safeguard Defenders, a human rights organization based in Madrid. We have spread the wrong claims, including the ones.
Laura Harth, a campaign director of Safeguard Defenders, said, “Researchers are” new multilinguals aimed at trusting organizations, intimidating, threatening, or doubts about their activities, or their activities. He is facing a attack. “
