麻省理工学院《科技评论》将于9月8日公布2026年度Innovators Under 35（35岁以下创新者）名单，共收录35位全球年轻科学家和工程师[1]。本次评选从550份提名中遴选而出[1]，旨在表彰这些创新者在科学与工程领域的突破性工作。
在最新一期的重点报道中，该杂志深入探讨了当前科技政策和安全领域的紧迫议题。美国国务院反外国虚假信息办公室将于2025年4月16日面临关闭[1]，引发外界对美国政府"审查工业综合体"政策方向的关注。与此同时，中国相关黑客对台湾进行了"前所未有"的AI攻击[1]，凸显了地缘政治风险。美国国防部冰上执法计划也宣布斥资最多2000万美元购买能释放电击的手套[1]，反映了执法技术的创新应用。
AI安全问题成为本期报道的另一核心焦点。Google DeepMind研究员Iason Gabriel指出："agents的悖论在于，让它们有用的东西——能够完成一系列任务——涉及放弃控制"[1]。AI Agent面临的主要风险包括误解目标、泄露敏感信息和遭受提示注入攻击等问题[1]。与此同时，Anthropic的Claude正在为AI文本和图像添加水印[1]，以应对虚假内容泛滥的挑战。
MIT Technology Review will unveil its 2026 Innovators Under 35 list on September 8, featuring 35 young scientists and engineers recognized for breakthrough work around the globe [1]. The publication received 550 nominations for this year's recognition program [1].
Beyond celebrating emerging talent, the latest edition addresses critical technology policy challenges, including the anticipated closure of the U.S. State Department's Office of Countering Foreign Disinformation on April 16, 2025 [1]. The publication also examines recent security threats, including unprecedented artificial intelligence attacks launched by Chinese-linked hackers against Taiwan [1], and reports that the U.S. Department of Defense has allocated up to 20 million dollars for an on-ice enforcement initiative to procure electrified gloves [1].
The coverage highlights pressing concerns around autonomous AI agents, with researchers documenting significant risks including goal misinterpretation, sensitive information disclosure, and prompt injection vulnerabilities [1]. Anthropic's Claude model is implementing watermarking technology for AI-generated text and images as a countermeasure [1]. According to Iason Gabriel, a researcher at Google DeepMind, "The paradox of agents is that the very thing that makes them useful—being able to accomplish a series of tasks—involves surrendering control" [1].
