HOOVER|StanfordStanfordUniversityWhitePaperIINSTITUTIONHuclatinteligenceUniversityJune2026Update:DeepseekAl and theGreat TalentCompetitionAmyZegartandEmersonJohnston
HAIStanfordStanfordUniversityUniversityArtficialIntelfgenceheGreatTalentCompetitionAuthorsEmerson Johnstonisa full-time research assistant at the Hoover Institution'sTechnologyPolicyAccelerator.Sheisagraduate ofStanford'sFordDorseyMaster's in International Policy and received her B.S. and B.A from Northeastern University,Institution, director of Hoover's Technology Policy Accelerator, and a senior fellowand associate director at Stanford University's Institute for Human-Centered AllHowtocitethispaperEmersonJohnstonandAmyZegart,"Update:DeepSeekAlandtheGreaTalent Competition"TechnologyPolicy Accelerator, HooverInstitution and the Stanford Institute for Human-Centered Artificial Intlligence, June 2026https:/doiorg/10.64576/0626
8Y55HAIStanfordStanfordUniversityArtificialInteligenceheGreatTalentCompetitionUniversityWhen DeepSeek Al released its R1paper in January 2025 hat rvealeda,"..a third ofthemostinteoralmodel rivaling OpenAl's ol forfarless compute, mediainarratives were dominated by two misconceptionscontributors to aChinese,First, that the Chinese company's breakthroughs were,model performing comparablythe result of Chinese researchers who could imitate,but not innovate; and second, that he research team,toOoenAl'solmodelnevertrained, published, or workedcomputer science field. Our 2025 report,"A Deep Pekinto DeepSeek AI's Talent and Implications for USin an international researchInnovation," extensively examined the 223 researchersenvironment."behind the company's five foundational papers andfound avery different picture.We found that DeepSeek didinnovate in significantChina'sdomesticAl talent pipeline is stillways, that its research team was far more experiencedgoing strongchan most assumed, and that more than halfoftheim had neverleft China for school or work. Together,.TheshareofDeenSeek AI researchersafiliatedinted to a domestic Al talent pipeline exclusively with Chinese institutions throughous their recorded careerssands at53.5percentrative allowed. They also sugested thatrateey built on the ssumption that China*Ten ofthe 31 Key Team members-researcherse talent or know-how to surpass the U.S. on its lisecdonaleven DepSeck Apapes pubishedin bela twoyars-ar lo liliaedecdusiveyown was already out of step with the evidencewih Chinse insitutions, meaning thid of Wearsued then that the United States needs tothe mostincegral contributors to a Chinese compete much more aggressively to attract, welcomemodelprforming comparably to OpenAl'soland retain the world's best and brightest while urgentlyimodel never trained, published,or workd in angrowing domestic capabilities by improvingK-12 andinternational research environment.higher education in STEM fields at home,New and better data reveals longer U.S.port asks how much has changed in the yealtrainingperiods,especiallyfortop talentWeextended our analysis f...
