Time cover on DeepSeek 'sweatshop' is AI-generated

The fabricated Time magazine cover was shared on X on February 10, 2025.

"DeepSeek Sweatshop, the dark side of innovation," its purported headline reads. 

Image

Screenshot of the false post, taken on February 18

DeepSeek's powerful R1 chatbot stunned investors and industry insiders. It also sparked panic on Wall Street with its ability to match the functions of its Western competitors at a fraction of the cost.

Analysts had long thought that the United States' critical advantage over China when it comes to producing high-powered chips -- and its ability to prevent the Asian power from accessing the technology -- would give it the edge in the AI race.

But DeepSeek researchers said they spent only $5.6 million and around 2,000 Nvidia chips developing the latest iteration of their model -- a fraction of the costs and computer power US tech giants have poured into AI (archived links here and here).

Governments from Rome to Seoul are also cracking down on the user-friendly Chinese app, saying they need to prevent potential leaks of sensitive information through generative AI services (archived link). 

The image was also shared elsewhere on X and in other languages, such as Slovak on Facebook and Russian on Telegram. 

'Not authentic'

But a spokesperson for the magazine told AFP that the doctored image "is not an authentic Time cover".

No such cover can be found in its archives of published covers (archived link).

Further checks also show the image contains a user handle from the Chinese app Xiaohongshu -- or RedNote in English.

Through the handle, AFP was able to trace the image to a post shared on February 8, 2025 (archived link).

The image is labelled a "fictional interpretation", with a caption that sarcastically says US Secretary of State Marco Rubio told Time magazine that the State Department "will classify DeepSeek a sweatshop" as the firm "exploits" its chips so it can develop an AI model using less computer power than US companies.

A keyword search on Google cannot find any genuine quotes from Rubio on DeepSeek in the RedNote post as of February 21.

Incoherent text

The fabricated Time cover has a different layout and contains random text without a specific published date compared to authentic ones.

Siwei Lyu, director of the University at Buffalo's Media Forensics Lab in the United States, told AFP the image did not come "from a real-world scene" and explained that the text does not form anything coherent because "AI models treat texts as visual contents instead of meaningful messages" (archived link).

Shu Hu, director of Machine Learning and Media Forensics Lab at Purdue University, said the "inconsistency in readability likely stems from the generation prompt lacking guidelines for maintaining consistent word style" (archived link).

Image

Screenshot comparison of the falsely shared Time magazine cover (L) and an authentic cover published on February 10, 2025 (R), with differences highlighted by AFP

AFP has previously debunked bogus Time covers here and here.