Why DeepSeek is Sparking TikTok-Like National Security Fears

THe quickly grows Chinese AI laboratory Deepseek It distorts concern for national security in the US, for fear of his AI model, the Chinese government can use the spying of American civilians, teach the ownership secrets and campaigns influence on salary. In his first briefing for journalists, print secretary of the Karolina Leavitt White House said that the national security council “studied” the potential security implications of Deepseek. This comes in the middle of the news that the US Navy has prohibited The use of deep duration among its ranks due to “potential security and ethical problems”.

Deepseek, which is currently on top of the Apple App Store in the USA, indicates the main point of the fold in Ai race of weapons between the US -ai China. Many years many leading technologists and political leaders They claimed that whatever the country developed and the fastest economic and military advantage over their rivals would have the fastest. Deepseek shows that Chinese AI developed much faster than many believed, despite this efforts from American policy creators to slowing down their progress.

However, other privacy experts claim that Deepseek’s data collection policies are not worse than those of his American competitors – and take care that these companies will be an excuse to call for a deregulation. In this way, a rhetorical battle over the danger of Deepseek is played on similar lines as in-lylbo Tictokwhich deeply divided the American public.

“There are completely valid questions of privacy and safety of data with Deepsek,” says Calli Schroeder, AI and Human Rights in Electronic Information Center (EPIC). “But everyone is present in us AI products.”

Read more:: What to know about deepseek

Concern about data

Deepseeek’s AI models act similarly to Chatgpt, Reply to user questions thanks to a huge amount of data and top processing capabilities. But his models are much cheaper to start: the company says that trained his R1 model To just $ 6 million, compared to the alleged tens of billions of dollars needed for the Openi -Is O1 model. Despite this huge bay, R1 still coincides with O1 on several mathematical and resonance.

Deepseek has built many open code resources, including the LLM V3, which rivals of OPENAI’s capacity of the GPT-4O closed source. Some care that by making such powerful technology open and repeatable, it provides the opportunity to people to use it more freely in malicious ways: creating bio-apoone, start large campaigns for phishing or fill in the Internet Silt backing. However, there is another contingent builder including the target VP and Main AI scientist Yann Lecunwho believe that the development of an open code is a more useful path for AI.

Another great concern is focused on data. Some privacy experts, such as Schroeder, claim that most of the LLM, including Deepsek, built on sensitive or defective databases: for example, information from the missing of data stolen biometry. David Sacks, AI and Krypto Emperor President Donald Trump, defendant Deepseek leaning against the output of Openi’s models to develop his own technology.

There is even more concern to use the user data to use Deepsek. The company’s privacy policy says it is automatic It collects a series of input data from its users, including IP and keys samples, and can use this to train its models. The personal data of the user are stored in the “safe servers found in the People’s Republic of China”, writes in politics.

For some Americans, this is particularly worrying because the generative Alans of the tools are often used in personal or high roles: to help their company strategies, manage finances or seek health advice. Such data can now be stored in a country with several data laws and a little transparency given how this data can be viewed or used. “It could happen that when the servers are physically accommodated in the country, it is much easier for them to approach them,” Schroeder says.

One of the main reasons why Tiktoc was banned in the US in the beginning of concern about how much information was collected by the Chinese home company of the application collected from Americans. If Americans start using Deepseek to manage their lives, privacy risks will be similar to “tictic on steroids,” says Douglas Schmidt, Dean of the Calculation School, Faculty of Faculty and Physics at William & Mary. “I think Tictok collected information, but these were mostly benign or generic data. But owners of large language models get a much deeper insight into the personality and interests and hope and dreams of users.”

Geopolitical concern

Deepseek is also alarming to those who see development as an existential weapon race between the USA -ai China. Some leaders claimed that Deepseek shows that China is now much closer to development Agi– Ai that can disperse at a human level or higher – than he had previously believed. American AI laboratories like Anthropically Some safety researchers work to relieve the damage of these increasingly fevern systems. But that’s unclear What a safety research team Deepseek uses. Cybersecurity of Deepseek’s models was also called into question. On Monday, the company limited new applications after saying that the application was targeted by a “big malicious attack”.

Before agi, a powerful, widely used Ai model could be influenced by the thought and ideology of its users around the world. Most AI models apply censorship in certain key ways or shows bias based on the information they are trained about. Users have found that Deepseek’s R1 refuses to answer questions about Massacre 1989 on Tiananmen Squareand hard That Taiwan is part of China. This caused the concern of some American leaders to use Deepsek to promote Chinese values ​​and political goals – either it was set up as a spying tool or cyber – Napade.

Read more: Artificial intelligence has sex and racial bias problems.

“This technology, if unverified, has the potential to feed the campaigns for disinformation, erodes confidence in the public and strengthen authoritarian narratives in our democracy,” Ross Burley wrote in a statement of resistance to Ross Burley Information Resistance, he wrote in a statement by email e -Long.

AI industry leaders and some Republican politicians responded by calling for mass investments in the US AI sector. President Trump said On Monday, that deepseek “should be a awake invitation for our industry that we need to be focused on the laser on a win competition.” Bags Posted on X that “Deepseek R1 shows that the Ai race will be very competitive and that President Trump was right to abolish Biden Eo,” inviting himself to AI Executive Command which, among other things, turned his attention to the potential short -term damage to the development of Ai too fast.

These fears could lead to now imposing stronger sanctions on Chinese technological companies or maybe even try to ban Deepseek alone. On Monday is the Committee to Select a House for the Communist Party of China called for stronger export control The technologies support Deepseek’s AI infrastructure.

But AI ethics are pushing back, claiming that the increase in Deepseek actually reveals an acute need for protective measures in the industry. “This has echoes of tictic bans: there are legitimate risks of privacy and security with the way these companies operate. But US companies that have been leading a lot of development of these technologies are similar to people’s data. Just because they do it in America does not make it better,” he says Ben Winters, AI Director and Data Privacy in America Consumer. “And Deepseek gives these companies another weapon in his chamber to say, ‘We really can’t regulate right now. “

As the ideological combat lines appear, Schroeder, in EPIC, warns users to be careful when using Deepseek or other LLM. “If you have concerns about the origin of the company,” he says, “be very, very careful in what you discover about yourself and others in these systems.”