HONG KONG -- DeepSeek, the Chinese tech startup that rivals OpenAI's ChatGPT, has been gaining crushed successful galore processing nations successful a inclination that could constrictive the spread of artificial quality adoption with precocious economies, a caller study suggested.
In the Thursday report, researchers from Microsoft said planetary adoption of generative AI tools reached 16.3% of the world’s colonisation successful the 3 months to December, up from 15.1% successful the erstwhile 3 months.
Yet the disagreement of AI adoption successful developed and processing countries is widening, the study noted, with AI adoption crossed precocious economies increasing astir doubly arsenic accelerated arsenic processing nations.
“We are seeing a disagreement and we are acrophobic that that disagreement volition proceed to widen,” said Juan Lavista Ferres, main information idiosyncratic for Microsoft's AI for Good Lab, which utilized anonymized “telemetry” to assistance way planetary instrumentality usage.
Countries that invested aboriginal and consistently successful integer infrastructure and AI led successful presumption of shares of users, including the United Arab Emirates, Singapore, France and Spain, according to the report. Some of Microsoft's figures overlapped with the findings of a Pew Research Center survey published successful October that mapped which countries are much excited than acrophobic astir AI. In some reports, for instance, South Korea stood retired successful its clasp of AI.
Microsoft has a vested involvement successful AI adoption — its concern and overmuch of the tech manufacture and banal marketplace is staking its aboriginal connected AI tools becoming much wide utilized and profitable — but Lavista Ferres said his laboratory is looking much broadly astatine the topic.
His researchers recovered that the emergence of Chinese startup DeepSeek, which was founded successful 2023, has fueled wider AI adoption crossed the processing satellite fixed its escaped and “open source” models – with cardinal components disposable for anyone to entree and modify.
When DeepSeek released its precocious reasoning AI exemplary called R1 successful January 2025, which it said was much cost-effective than OpenAI’s akin model, it raised eyebrows successful the planetary exertion manufacture and galore were amazed by however China is catching up with the U.S. successful technological advancements. Leading subject diary Nature published peer-reviewed probe co-authored by DeepSeek laminitis Liang Wenfeng successful September, describing it arsenic a “landmark paper” from the Chinese startup.
Lavista Ferres said DeepSeek is simply a “good model” for tasks similar mathematics oregon coding, but it operates otherwise from U.S.-based models connected topics similar politics.
“We person observed that for definite benignant of questions, of course, they travel the aforesaid benignant of entree to the net that China has,” helium said. “Which means that determination volition beryllium questions that volition beryllium answered precise differently, peculiarly governmental questions. In galore ways that tin person an power connected the world.”
DeepSeek offers a free‑to‑use chatbot connected web and mobile, and has besides fixed developers planetary entree to modify and physique connected its halfway engine. Its deficiency of subscription fees has "lowered the obstruction for millions of users, particularly successful price‑sensitive regions,” Microsoft's study said.
DeepSeek didn’t instantly respond to a petition for remark connected the report.
“This operation of openness and affordability allowed DeepSeek to summation traction successful markets underserved by Western AI platforms,” the study added. “DeepSeek’s emergence shows that planetary AI adoption is shaped arsenic overmuch by entree and availability arsenic by exemplary quality.”
Developed countries including Australia, Germany and the U.S. person sought to bounds the usage of DeepSeek implicit alleged information risks. Microsoft past twelvemonth banned its ain employees from utilizing DeepSeek. Adoption of DeepSeek remained debased successful North America and Europe, the study found, but it surged successful its location state China, arsenic good arsenic Russia, Iran, Cuba, Belarus – places wherever U.S. services look restrictions oregon wherever overseas tech entree is limited.
In galore places, DeepSeek’s prevalence correlated with it being a default chatbot connected wide disposable phones made by Chinese tech companies similar Huawei.
DeepSeek’s marketplace stock successful China was 89%, the study estimated. That’s followed by Belarus’s 56% and Cuba’s 49%, some of which besides had debased AI adoption much broadly. In Russia, its marketplace stock was astir 43%.
In Syria and Iran, DeepSeek’s marketplace stock reached astir 23% and 25%, respectively, the study added. In galore African countries including Ethiopia, Zimbabwe, Uganda and Niger, DeepSeek’s marketplace stock was betwixt 11% to 14%.
“Open‑source AI tin relation arsenic a geopolitical instrument, extending Chinese power successful areas wherever Western platforms cannot easy operate,” the study said.
___
O'Brien reported from Providence, Rhode Island.
English (CA) · 
English (US) · 
Spanish (MX) ·