Google员工因军事AI合同争议而离职

Google DeepMind的多名员工因公司与美国和以色列军方的AI技术合作而宣布离职。1研究工程师Sofia Liguori本周从Google DeepMind辞职，表示公司在军事行动中推进AI系统令她感到羞愧。1AI研究员Andreas Kirsch也在本月早些时候因类似原因离职，他指出"Google的治理和监督不足以在军事和机密环境中安全和负责任地部署AI"。1

Google与美国国防部的合同允许AI技术用于"任何合法的政府目的"，但规定"不应用于国内大规模监视或自主武器（包括目标选择）而没有适当的人工监督和控制"。1除Google外，OpenAI、Amazon、Microsoft、SpaceX、Oracle、Nvidia和Reflection等多家科技公司也签署了类似合同。1其中，Google与Amazon的Project Nimbus合同价值12亿美元，为以色列政府和军方提供AI技术和云计算基础设施。1

员工和专家对这些AI军事应用的充分性和人工监督存在广泛担忧。退役中将Jack Shanahan表示，军方的哲学似乎是"尽快部署AI技术，尽可能广泛部署，可能以测试和评估的严谨性和纪律为代价"。1AI专家Sarah Shoker指出，"AI压缩了操作员做决定的时间，在高压环境中风险特别大"。1此外，离职员工表示对批评公司感到恐惧，Liguori称"原则上我一生都不被允许批评Google"。1

Multiple Google DeepMind employees have departed the company in protest of its artificial intelligence agreements with the U.S. military and Israeli government.1 Research engineer Sofia Liguori resigned this week, stating that the company's advancement of AI systems for military operations filled her with shame.1 AI researcher Andreas Kirsch also left earlier this month, expressing concern that "Google's governance and oversight are insufficient to safely and responsibly deploy AI in military and classified environments."1

Google's contract with the Pentagon permits AI technology to be used for "any lawful government purpose," though it stipulates the technology "should not be used for domestic mass surveillance or autonomous weapons, including target selection, without appropriate human oversight and control."1 The company has also partnered with Amazon on Project Nimbus, a $1.2 billion deal providing AI technology and cloud computing infrastructure to the Israeli government and military.1 Google is among numerous technology firms—including OpenAI, Amazon, Microsoft, SpaceX, Oracle, Nvidia, and Reflection—that have signed similar military contracts.1

Concerns about the deployment model have been raised by military and technology experts alike. Retired Lieutenant General Jack Shanahan indicated that military philosophy appears to be "to deploy AI technology as quickly as possible, as broadly as possible, potentially at the expense of the rigor and discipline of testing and evaluation."1 Technology policy researcher Sarah Shoker pointed out that "AI compresses the time operators have to make decisions, which is particularly risky in high-pressure environments."1 Among departing employees, there is also apprehension about speaking publicly. Liguori noted that "in principle, I am not allowed to criticize Google for the rest of my life."1