美国军方一项针对中国船舶的武装行动因AI系统生成的虚假情报险些被执行1。据CNN于周五报道，这一事件发生在春季1。特种作战司令部的分析师使用AI聊天机器人综合开源数据和机密信号情报，但该系统误识别了船舶货物清单，错误地声称该船舶携带核武器项目零部件1。分析师随后又使用AI工具将这一虚假结论格式化为官方报告并传递至指挥链1。直到军机已经升空后，问题才被发现，相关操作被紧急中止1。
这一事件揭示了军方在采用AI加速决策时面临的核心风险1。GovAI研究员Jake Steckler指出："LLM的不确定性对任何可能导致使用武力的决策尤其关键，这些决策关乎生死"1。五角大楼虽然将AI描述为能显著加快"杀伤链"速度的工具，但同样的速度也可能使幻觉在人工监督不足的情况下扩散1。
A U.S. Special Operations Command analyst's use of an AI chatbot to process intelligence nearly resulted in an armed military action against a Chinese vessel, according to reporting by CNN on Friday 1. The artificial intelligence tool generated false information claiming the ship carried components for nuclear weapons programs, based on its analysis of open-source data and classified signals intelligence 1. The analyst used the AI tool twice: first to synthesize the data, and then again to format the erroneous conclusion into an official report, which was then passed up the chain of command 1.
The operation proceeded to the point where military aircraft had already been launched before the problem was discovered and the action was hastily halted 1. Jake Steckler, a researcher at GovAI, emphasized that "the uncertainty of large language models is especially critical for any decision that could lead to the use of force, decisions about life and death" 1. While the Pentagon has characterized AI as capable of significantly speeding up the "kill chain," that same acceleration creates risks that hallucinations could spread without adequate human oversight 1.
评论
还没有评论，欢迎留下第一条。