Project Maven 由早期引发超过 3,000 名 Google 员工在 2018 年抗议，转变为美军 AI 战争的核心系统。其 Maven Smart System 将感测器、地图、AI 侦测与目标流程整合，Vice Admiral Frank Whitworth 也从怀疑者转为支持者。国会先前已投入 10 亿美元，系统之后又取得 4.8 亿美元、1 亿美元与上限 13 亿美元的合约；英国据报也考虑约 7.5 亿英镑（约 10 亿美元）协议。

数字显示其扩张速度惊人。到 2025 年，约有 25,000 名美方人员使用 Maven，至少 32 家公司参与，使用量较 1 月翻倍以上，资料库累积 10 亿笔 AI 侦测，侦测速度接近先前的 5 倍。CENTCOM 汇入 179 路即时资料流，拥有 13,000 个帐户与 2,500 名常态使用者；Maven 在两年间进入 141 场演习与实验、130 多个地点，并把部分打击节奏由每天不到 100 个目标推升至 1,000 个，在整合 LLM 后又升至每天 5,000 个目标。

但文章核心是风险与责任，而不只是效率。Maven 已用于也门、红海、伊拉克、叙利亚、伊朗相关行动，以及边境与海上拦截；单一目标单位可把从感测到开火的时间由数小时压缩到数分钟。批评者指出幻觉、错误资料、频宽暴露、缺乏正式 doctrine 与训练，并质疑这其实已是 weapons system；支持者则主张战场速度与规模若无 AI 无法达成。文章因此把 Maven 描写为一场已被制度化的运动：美国正把「适当的人类判断」留在程序中，却同时更快地把致命决策推向自动化。

Project Maven evolved from a Pentagon experiment that triggered protests by more than 3,000 Google employees in 2018 into a central system for US AI warfare. Its Maven Smart System fuses sensors, maps, AI detections, and targeting workflows, and even Vice Admiral Frank Whitworth shifted from skeptic to advocate. Congress had already spent $1 billion on Maven, and the platform later added contracts worth $480 million, $100 million, and a ceiling of $1.3 billion; the UK was also reportedly considering a deal worth about £750 million ($1 billion).

The numerical scale of expansion is striking. By 2025, roughly 25,000 US personnel were using Maven, at least 32 companies were involved, usage had more than doubled since January, the data store held 1 billion AI detections, and detection speed was nearly five times faster than before. CENTCOM alone fed 179 live data streams into the system and had 13,000 accounts with 2,500 regular users; across two years, Maven appeared in 141 exercises and experiments and more than 130 sites, while parts of the strike cycle rose from under 100 targets per day to 1,000, and then to 5,000 per day after LLM integration.

But the article’s core issue is not efficiency alone; it is risk, accountability, and the delegation of lethal judgment. Maven was used in operations tied to Yemen, the Red Sea, Iraq, Syria, Iran-related conflict, and border and maritime interdiction, with one targeting cell cutting the time from sensing to shooting from hours to minutes. Critics warn about hallucinations, bad data, bandwidth exposure, and the lack of formal doctrine and training, arguing that the system already functions as a weapons system; supporters argue that modern combat speed and scale are impossible without AI. The result is a portrait of Maven as an institutionalized movement in which the United States keeps “human judgment” in the loop while moving lethal decision-making steadily closer to automation.