根据中国公开研究文件与英国军情五处（MI5）发布的间谍警报显示，中国知名人工智慧新创公司 Z.AI（智谱 AI）的四位创办人曾参与一项由中国通用技术研究院（CGTRI）与中国国家自然科学基金委员会共同资助的联合 AI 研究项目。英国军情五处指控 CGTRI 与中国国家安全部（MSS）有极为密切的联系，其主要目的是资助能够直接提升中国间谍技术能力的学术研究，并警告任何机构或个人若继续进行由该院资助的研究将可能面临刑事后果。对此，中国驻英与驻美使馆发言人皆驳斥相关指控毫无根据，强调中国反对歪曲事实的抹黑行为，而 Z.AI 与相关创办人则未对置评请求作出回应。
该项为期近四年的研究计划于 2018 年 1 月至 2021 年 12 月间进行，总预算达 294 万人民币，时间点与 Z.AI 在 2019 年成立初期重叠。研究报告指出，该计划由清华大学、东南大学与 CGTRI 共同执行，旨在开发能分析社群媒体贴文、识别重大事件并绘制人际关系网络的 AI 工具，其核心目标之一是建立「涉华重点人物」知识库。Z.AI 共同创办人兼清华大学教授李涓子与 CGTRI 工程师曹伟兵等人共同主导了整体技术方案，而执行长张鹏、唐杰与许斌等另外三位创办人则名列论文合著者、专利共同发明人或软体著作权人，五位核心创办人中仅董事长刘德兵未出现在该项目纪录中。
Z.AI 源自清华大学的研究成果，曾获得中国地方政府及阿里巴巴、腾讯等科技巨头支持，并于 2026 年 1 月在香港上市，目前市值约 320 亿美元。然而，该公司近期面临多项美国与西方国家的审查与指控，包括因涉嫌协助军方研究而被美国限制取得技术、被 OpenAI 与 Anthropic 指控透过「蒸馏」技术系统性提取美国模型数据，以及处理软体漏洞的方式引发国安疑虑。咨询公司 Garnaut Global 的报告指出，Z.AI 的模型在微软、英特尔与 Google 等西方企业软体中发现了 2,436 个漏洞，却仅向受影响厂商通报 5%，随后更撤下公开帐本，改由包含中国国安部运作的数个政府资料库来管理这些漏洞。
According to publicly available Chinese research papers and an espionage alert from the UK's MI5 intelligence service, four founders of the prominent Chinese artificial intelligence startup Z.AI (also known as Zhipu) participated in a joint AI research project co-funded by the China General Technology Research Institute (CGTRI) and the National Natural Science Foundation of China. MI5 accused CGTRI of having very strong ties to China's Ministry of State Security (MSS) and existing primarily to fund academic research that directly enhances the spy agency's technical espionage capabilities, warning of potential criminal consequences for anyone continuing CGTRI-funded work. Spokespeople for China's embassies in the UK and Washington dismissed the allegations as baseless and opposed attempts to distort facts, while Z.AI and the involved founders did not respond to requests for comment.
Running for nearly four years from January 2018 through December 2021 with an approved budget of CNY2.94 million, the research project overlapped with Z.AI's founding in 2019. According to a 78-page report, the initiative was jointly conducted by Tsinghua University, Southeast University, and CGTRI to develop AI tools capable of analyzing social media posts, spotting key events, and mapping interpersonal connections, with a core goal of building a knowledge base of China-related key persons. Z.AI co-founder and Tsinghua University professor Li Juanzi co-led the overall technical plan alongside CGTRI engineer Cao Weibing, while three other Z.AI founders—CEO Zhang Peng, Tang Jie, and Xu Bin—contributed as paper co-authors, patent co-inventors, or software copyright holders, leaving Chairman Liu Debing as the only key founder not named in the project documentation.
Spun out of Tsinghua University and backed by local Chinese governments as well as tech giants Alibaba and Tencent, Z.AI went public in Hong Kong in January 2026 and currently holds a market capitalization of roughly $32 billion. Nevertheless, the company faces intensifying Western scrutiny, including US export restrictions over alleged military AI assistance, accusations from OpenAI and Anthropic of systematically extracting proprietary knowledge from American models via distillation, and alarms over its handling of software vulnerabilities. A report by consultancy Garnaut Global revealed that Z.AI discovered 2,436 software flaws across major Western firms like Microsoft, Intel, and Google, but disclosed only 5% to the affected vendors before taking down its online ledger and handing management of the vulnerabilities over to Chinese government databases, including one run by the Ministry of State Security.