Category
Latest news

Russian State TV Claims China’s DeepSeek AI Is Built on Soviet-Era Code From 1985

3 min read
Authors
Russian State TV Claims China’s DeepSeek AI Is Built on Soviet-Era Code From 1985
The DeepSeek app on a phone on January 28, 2025, in Hong Kong, China. (Source: Getty Images)

Russian state television aired a report claiming that China’s advanced AI chatbot, DeepSeek, was developed using Soviet-era programming code from 1985, The Moscow Times reported on February 3.

The claim, broadcasted on the Russia-1 network’s ‘Vesti’ program, was entirely based on a satirical article from the parody news outlet Panorama.

The fake story originated on January 29, when Panorama published a fictional interview with DeepSeek’s developer, Lian Wenfeng. The article falsely quoted Wenfeng as saying that DeepSeek was built on the foundation of OGAS , a Soviet-era computer system designed by academician Viktor Glushkov and that without this technology, China’s AI “would never have caught up with the Americans and their ChatGPT.”

Despite the website’s clear disclaimer that all of its stories are “grotesque parodies of reality”, Russian state media failed to verify the information and presented the claim as a fact.

The story quickly spread across regional Russian news outlets, including DonPress and Lenta Novostey Kurska. It was also shared by former Russian media Regnum editor-in-chief Yuri Baranchik before making its way onto national television.

On February 2, Russia-1 aired the fabricated claim during its 11:00 am broadcast. The segment did not question the authenticity of the information, instead presenting it as a breakthrough revelation.

Russian state television broadcast, airing the claim that Chinese AI was created using ancient Soviet code from the mid-'80s. (Source: Russia 1)
Russian state television broadcast, airing the claim that Chinese AI was created using ancient Soviet code from the mid-'80s. (Source: Russia 1)

The controversy comes as DeepSeek has been making headlines worldwide. On January 27, it surpassed OpenAI’s ChatGPT in the US App Store’s ranking for most downloaded free applications.

The AI model has demonstrated capabilities comparable to leading Silicon Valley competitors, and in some areas, it has even outperformed them.

In December, reports revealed that training DeepSeek-V3 cost less than $6 million in computing power, significantly lower than its Western rivals. However, recent reports state that estimated DeepSeek’s total hardware investment surpasses $500 million, far exceeding the publicly stated $6 million pre-training cost.

Following the chatbot’s rapid ascent, shares of major Western tech firms took a hit. On a single day, Nvidia’s stock plummeted by over 17% on the NASDAQ, erasing approximately $620 billion from its market capitalization and knocking it from its position as the world’s most valuable company, behind Apple and Microsoft.

Amid DeepSeek’s success, questions have emerged about the origins of its technology. Microsoft, OpenAI’s key investor and technology partner, raised concerns that DeepSeek may have incorporated proprietary elements of ChatGPT. The company has launched an internal investigation into possible intellectual property violations.

David Sacks, a senior AI advisor in President Donald Trump’s administration said there’s “substantial evidence” that DeepSeek distilled knowledge out of OpenAI’s model to develop its AI. He pointed specifically to distillation techniques, where one AI model learns from another to replicate its capabilities.

Earlier, the Russian hacker group named 22c and a cyber operator known as PalachPro claimed responsibility for the recent global failure of ChatGPT.

See all

OGAS, or “National Automated System for Computation and Information Processing” was a Soviet project to create a nationwide information network.