Bloomberg 记者 Michael Shepard、Rebecca Torrence 与 Edward Ludlow 报导，Palantir CTO Shyam Sankar 将伊朗战争定性为可能第一个主要由人工智能驱动的冲突，也是现代战争的新分野。Sankar 在华府 Hill and Valley Forum 上表示，这场作战正在展现科技对实际作战效率的影响，而非停留在概念层面。
文章指出，美军在伊朗战场大量使用 Palantir 的 Maven Smart System，这个数位任务控制平台可整合大量即时资料以加快决策速度。该系统近期被国防部门列为 program of record，预期意味著更稳定的资金与持续研发。Sankar 认为目标选取并非单纯「开火」印象，而是大规模流程化作战，类似可缩短原本数月时程的供应链管理作业。
文中也提到 AI 在目标选择上正成为监督焦点。Palantir 在 Maven 中使用 Anthropic 的 Claude，但美国国防部本月将 Anthropic 列为供应链风险，可能迫使 Palantir 寻找替代供应商；Anthropic 已提起联邦法院诉讼，听证会定于次周二。Sankar 另批评美国顶尖实验室过度著迷 AGI，并以第二次世界大战为例说明：共建造154种空战机构型，只有约10种关键，比例约6.4%，他认为可仿效达尔文式实验方式，同步部署154种低成本 LUCAS 无人机变体。
On March 25, 2026 at 3:07 AM GMT+8, Bloomberg reporters Michael Shepard, Rebecca Torrence, and Edward Ludlow reported that Palantir CTO Shyam Sankar characterized the Iran war as likely the first major conflict driven by artificial intelligence, and a turning point in modern warfare. Speaking at the Hill and Valley Forum in Washington, Sankar said the campaign is increasingly showing technology’s practical impact on combat efficiency rather than remaining a conceptual issue.
The article says U.S. forces are using Palantir’s Maven Smart System at scale in Iran, a digital mission-control platform that integrates large volumes of real-time data to speed decision-making. The system was recently designated a program of record by defense officials, which is expected to imply more stable funding and continuous development. Sankar argued that target selection is not simply the impression of “shooting” but a large-scale, procedural operation, similar to supply-chain management processes that can compress timelines that used to take months.
The role of AI in target selection is becoming a key oversight focus. Palantir uses Anthropic’s Claude in Maven, but the U.S. Department of Defense named Anthropic a supply-chain risk earlier this month, which may force Palantir to seek alternate suppliers; Anthropic has filed federal litigation, with a hearing set for the following Tuesday. Sankar also criticized U.S. top AI labs for over-focusing on AGI, and cited the World War II example: 154 airframe types were built and only about 10 were key, about 6.4%; he suggested a Darwinian approach could justify fielding 154 low-cost LUCAS drone variants simultaneously.