China: DeepSeek’s AI Breakthrough
Published: June 5, 2025
This is the first edition of our News Breakdown weekly email! Find out more
In January 2025, a little-known startup in Hangzhou, China launched DeepSeek AI chatbot.
DeepSeek offered a reasoning AI, rivalling the best model at the time (OpenAI’s o1) at a fraction of the cost. It became the most downloaded app that month.
DeepSeek created uncertainty in the AI industry, causing a drop in market value of many US tech companies.
In this report, we unpack how the media reacted to DeepSeek’s launch and follow up on its consequences.
Fox News: US-based conservative news outlet
Republican: right-leaning bias, most commentators support Republican causes.
Most-watched cable TV news network in the US.
Controversial: involved in scandals over politically motivated misleading claims.
Angle: DeepSeek as a national security threat to the United States.
The article focuses on US politics, written for a mostly American readership.
It describes DeepSeek’s launch as a “fallout” in the headline, suggesting it was a negative event.
Fox presents the news in a way that is favoured by Republican politicians and voters, who tend to be harder on China than Democrats.
The article focuses on a call from a Republican Senator to ban all AI cooperation with China due to concerns that the app could be used to spy on US users.
The act included a complete ban on any imports of Chinese AI technology, and on exports of any AI tech and intellectual property to China.
The proposed law has gained little support so far, as Senator Hawley is still its only sponsor.
However, on Feb 7, 2025, another bill was proposed to the US House of Representatives, sponsored by 14 Republican and 18 Democrat representatives, looking to ban DeepSeek from any government devices (except for law enforcement purposes).
The bill did not yet pass the vote in the House of Representatives. It would then need to be approved by the Senate and signed by the President.
DeepSeek is already banned on all state-owned devices in New York, Texas and Virginia.
The Fox News article included appeals to readers’ emotions, such as with this quote from Senator Hawley:
“Every dollar and gig of data that flows into Chinese AI are dollars and data that will ultimately be used against the United States.”
It is true that DeepSeek sends users’ data onto its servers in China.
However, third-party services like Perplexity can use DeepSeek models hosted in Europe or the US, without sending data to China.
Under China’s cybersecurity law, the Chinese government can access any data stored in the country when needed.
DeepSeek has also shared its data with some companies flagged as security threats by the US.
This perspective assumes that China will use the data collected on US users by Chinese firms in some form of aggression against the United States, which could be reasonable in the context of an ongoing trade war.
Fox News article concludes by mentioning that President Biden “slapped export controls of AI chips” in “his final week in office”.
This is misleading, as Biden’s administration already introduced AI chip export restrictions earlier, including in 2022.
Al Jazeera: international broadcaster based in Qatar
Alternative: Focuses on issues affecting the Global South.
International: Global focus, broadcasts in English and Arabic.
Partially state-funded by the government of Qatar, but editorially independent.
Angle: DeepSeek as a cost-effective AI innovation and a business disruptor.
Al Jazeera’s article is neutral and industry-focused, but avoids the topics of data privacy, censorship or national security.
The article’s main points included the cost-effectiveness of DeepSeek compared to its rivals, as well as the fall in Nvidia stock prices caused by the app’s launch.
The article reports that creators of DeepSeek claim to have trained their models for under $6 million – by comparison, Open AI’s GPT-4 cost over $100 million to train.
However, this figure is believed to exclude at least $10 million in other expenses of training the model.
The article mentions that DeepSeek’s owner had stockpiled 10,000 advanced A100 chips (used for AI training) before the US introduced export restrictions on this model.
AJ clarifies that DeepSeek claims to have neither used the A100, nor the other advanced H100 chips, in training its model, except using less advanced H800 chips.
In the Global South, the West’s technological monopoly is often viewed with scepticism and China is seen as a practical alternative.
Qatar-China relations: Al Jazeera, though technically independent, is partially funded by the Qatari government, which has close economic and diplomatic ties with China.
In 2019, Al Jazeera ran a series of pieces on Chinese investment in Africa that emphasised economic benefits and downplayed criticism from Western economists or NGOs.
Qatar and China also have a mutual non-interference policy, affecting Al Jazeera’s editorial angle, particularly on politically sensitive topics.
Daily Mail: UK’s most read printed newspaper
Conservative: openly endorses the UK Conservative party since 1945.
Tabloid: Uses sensationalist language to drive engagement.
Angle: DeepSeek as a potential AI weapon developed by China.
In the headline, DeepSeek is called a “secret AI weapon for WW3”, drawing the reader’s attention with a scary speculation about a global war.
The article mentions ways in which DeepSeek can be a security threat to the UK and Western countries:
Disinformation – pointing to DeepSeek spreading false information on events that are politically sensitive in China. Daily Mail calls Taiwan a “potential flashpoint for future world wars”.
Spying - the article claims that (according to unnamed “experts”) DeepSeek “will be able to collect information on users' devices, typing patterns, IP address and more”, adding that this information could then be passed onto the Chinese government.
The article reminds the readers of China’s Military-Civil Fusion: a doctrine requiring private businesses to support the government's efforts when needed.
The article uses hyperbolised language and adds disclaimers about the uncertainty of these claims. This allows the outlet to potentially deny making any specific claims.
In the past, the Daily Mail has used similar rhetoric when referring to TikTok, framing it as a youth-targeted surveillance tool.
In the headline, Daily Mail claims that China “stole” tech advantage from the US, blaming President Biden for not preventing this.
This equates DeepSeek and China, promoting the narrative of US-China competition, rather than seeing it as a competition between private companies.
The article also uses this as an opportunity to present the Biden administration as weak, catering to a largely pro-Trump audience.
Daily Mail also suggests that Biden administration banned the exports of the best chips used for AI training too late (in 2022).
The article reports that DeepSeek’s founder had controlled a cluster of 10,000 top-range A100 chips in the US in 2022, suggesting they were used to train DeepSeek, and criticising Biden’s export policy.
Independent analysts confirm that DeepSeek has access to 10,000 A100 chips. As the article notes, DeepSeek reports that it used other, less advanced chips that are allowed for exports to China.
China Daily: China’s largest English-language newspaper
Subject to censorship by the Chinese Communist Party.
Historically seen as more liberal than other Chinese state news outlets.
Angle: DeepSeek as a globally recognised Chinese AI breakthrough.
The article portrayed DeepSeek as a highly praised app, citing a “flurry of positive reactions” from its Western rivals due to its cost effectiveness and efficiency.
The article selected quotes from the CEOs of Microsoft, Meta and Google, praising DeepSeek for its “genuine innovations” and contribution to AI advancement.
However, Microsoft have also banned the use of DeepSeek among their employees due to data privacy and propaganda concerns in May 2025.
China Daily then shares quotes from the CEOs of DeepSeek’s competitors in the AI sector – OpenAI and Perplexity, also praising the “breakthrough”.
It follows up by quoting reports from MIT University (US) and KPMG consulting group (UK), backing up the impressiveness of the advancement.
The article avoided discussion of censorship, surveillance, or ethical concerns.
China Daily’s coverage is consistent with the government's messaging that prioritises state success, self-reliance and international prestige.
It wraps up with a future outlook: “Industry insiders said this development could level the playing field between large tech companies and smaller startups, potentially fostering more collaboration and innovation in the AI sector.”
BBC: UK’s publicly funded national broadcaster
Has a legal commitment to impartiality
Funded through a TV licence fee, paid annually by all UK households that watch or record live TV
Angle: a balanced analysis of DeepSeek‘s breakthrough.
BBC offered a balanced and cautious take on the DeepSeek R1’s release.
The article both acknowledges the system’s capabilities and raises concerns about content censorship, data privacy and transparency.
BBC quotes Donald Trump's description of DeepSeek as a "wake-up call [for American companies]”.
The article discussed the cost-efficiencies. However, the BBC was cautious with its statements, clarifying that the developers of DeepSeek “claim” that the model cost less than $6 million to train.
The article also cited 1989 Tiananmen Square massacre (when at least hundreds of civilian protesters were killed) as an example of DeepSeek having been trained to avoid politically sensitive questions.
BBC discussed broader geopolitical implications, including responses from Australia and Italy — both countries that have taken steps to block the use of DeepSeek over the concerns that the data is being stored in unsafe “China-based servers”.
As a publicly funded broadcaster, the BBC must keep neutrality, however its coverage is still reflective of British liberal-democratic norms and at times displays left-leaning bias.
DeepSeek openly published its model’s parameters and its research breakthroughs in AI efficiency, allowing others to benefit from the advancement.
This helped a number of Chinese AI companies, like Alibaba and Tencent, to catch up with US competitors.
Soon after launch, DeepSeek was targeted by a large-scale cyberattack that caused it to pause registering new users in China and led to the loss of some sensitive data.
In response to competition from DeepSeek, Google and OpenAI introduced lower pricing tiers and cheaper models.
Other companies, like Microsoft and Perplexity, incorporated DeepSeek models into their products.
However, DeepSeek has increased its prices and removed promotional discounts over time, leading to some costs rising by 300% since.
Some countries banned the use of DeepSeek: Italy, India, Australia, South Korea and Taiwan. Some governments and organisations also introduced restrictions or bans.
Nvidia’s stock price recovered 70% of its loss caused by DeepSeek after a month. However, it then fell even further due to the US escalating a trade war.
By June 2025, Nvidia’s stock price is again approaching its pre-DeepSeek level.
Authors Anton Kutuzov, Tamara Krivskaya
This was the first edition of News Breakdown
To continue receiving these reports, please choose you subscription below!
P.S. We really need your support! The revenue would allow us to stay independent long-term and keep our content ad-free. Your membership means everything to us.
Right now, get a month of free trial with any option: