Investigation of China’s DeepSeek finds America losing edge in tech talent

America looks to be losing its advantage in the top tech talent needed to compete globally for advanced artificial intelligence.

A Stanford University investigation into gains by DeepSeek published this week revealed that the Chinese firm’s sudden rise came from homegrown researchers.

DeepSeek’s success in creating a powerful model that compares favorably to Western AI models while relying on Chinese talent is an “early-warning indicator” for America, according to Amy Zegart, senior fellow at Stanford’s Hoover Institution. 

“More than half of [DeepSeek researchers] never left China for schooling or work, demonstrating the country’s growing capacity to develop world-class AI talent through an entirely domestic pipeline,” Ms. Zegart wrote in a paper with researcher Emerson Johnson. “And while nearly a quarter of DeepSeek researchers gained some experience at U.S. institutions during their careers, most returned to China, creating a one-way knowledge transfer that benefits China’s AI ecosystem.”

Skepticism about DeepSeek’s rapid rise has swirled since it unveiled a powerful model this year. Western AI experts have viewed the Chinese firm’s low-cost claims with suspicion, and Commerce Secretary Howard Lutnick is among those who have accused DeepSeek of stealing American intellectual property. 

Suggestions that DeepSeek’s emergence this year was built by an inexperienced team relying entirely on others’ work appear misplaced, according to Stanford’s investigation of the more than 200 DeepSeek researchers’ published work. 

“While OpenAI continues to receive global recognition, many of DeepSeek’s central contributors — at least by traditional bibliometric standards — were better published, more consistently cited, and arguably more academically established at the time of their breakthrough,” the Stanford report said.

China’s advances in the talent needed for AI research have been noticed by the national security community. The Center for a New American Security published a paper this week arguing that the U.S. must get “more aggressive” in pursuing a lead over China in AI via the metrics of computer capacity, data, institutions and talent. 

“Bolstering the United States’ own AI talent base through enhanced immigration channels could go a long way to securing American advantage, as could weaponized brain drain of China’s top AI researchers — as long as their work can be appropriately directed toward nonsensitive research areas,” the CNAS report said. 

Concern about America’s own tech expertise transferring to China and DeepSeek’s close connections with the Chinese government have permeated Capitol Hill. 

Earlier this month, the House Select Committee on the Chinese Communist Party published the findings of its probe into DeepSeek and determined that the Chinese firm violated U.S. companies’ terms of service and had funneled Americans’ data to China. 

“DeepSeek personnel infiltrated U.S. AI models and fraudulently evaded protective measures under aliases and purchased dozens of accounts using a sophisticated network of international banking channels,” the committee said in a report on its findings. “This allowed DeepSeek personnel to mask their identities, conceal their transactions, and avoid detection.”

The Stanford researchers believe DeepSeek’s research network presents new problems that can’t be solved through protection against theft and espionage alone.

“Top global talent has options. DeepSeek’s talent story suggests that the United States cannot assume a permanent talent lead,” the Stanford investigation said. “Instead, the nation needs to compete much more aggressively to attract, welcome, and retain the world’s best and brightest while urgently growing domestic capabilities by improving K-12 STEM (science, technology, engineering, mathematics) education at home.”