2026年5月29日 / 美国东部时间上午9:52 / CBS新闻
摩洛哥丹吉尔——本月，在撒哈拉沙漠与大西洋沿岸交汇的摩洛哥南部干旱地区，沙漠的宁静被爆炸的轰鸣声和密集的枪炮声打破。
美军在参与多国“非洲狮2026”军事演习期间，常规火炮产生的硝烟弥漫在空中。但美国陆军同时也利用这场战争演习，测试了一系列由人工智能驱动的系统。
“非洲狮”演习是美国主导的非洲规模最大的军事演习，与30个伙伴国共同开展，旨在为未来战争进行演练。而未来战争越来越离不开人工智能。
除军方部队外，十多家私营国防承包商展示了相关产品，并直接从士兵那里获取反馈，争相争夺帮助美军现代化的角色和合同。
缩短“杀伤链”
当士兵们演练传统战场战术时，一辆顶部搭载机枪的机器人悄然驶过摩洛哥沙漠。无人机在附近升空携带着炸药，另一架原型四轴飞行器则搭载了9毫米步枪。
演习中展示的人工智能主要应用之一，是致力于缩短“杀伤链”——即从识别目标到扣动扳机所需的一系列使用致命武力的行动步骤。
美国陆军中校拉蒙·隆格雷罗对CBS新闻表示，位于距模拟战场数百英里外阿加迪尔联合作战中心的人员，使用了美国国防科技公司Palantir开发的人工智能驱动平台，“以实现比常规更快的决策循环”。
美国军事人员在摩洛哥阿加迪尔的联合作战中心开展工作，此次“非洲狮2026”联合演习使用了Palantir公司开发的人工智能驱动平台。CBS新闻
“五年前，这类决策可能需要两到三个小时，”他在谈及某次演习中的决策时对CBS新闻表示，“而我们仅用了三分钟。”
在那次演练中，杀伤链的最终环节仍由人类负责批准目标并命令火炮部队实施打击。但隆格雷罗告诉CBS新闻，无需人类介入即可自主决定何时扣动扳机的系统已经存在，以此进一步节省时间。他不愿透露是否有任何实战行动使用过此类系统。
在作战中心，数十人坐在大屏幕前协调地面行动。
支撑此次大部分行动的系统是“马文项目”（Project Maven），这是五角大楼的旗舰人工智能项目，由Palantir公司开发。马文项目会接收海量战场数据，并利用人工智能识别模式、为指挥官优先处理信息，例如确定打击目标。
尽管与五角大楼产生争执，Anthropic的Claude AI仍至关重要
据多名熟悉演习中使用的系统的军方和行业消息人士透露，马文项目与人类操作员的交互依赖于Anthropic公司的Claude大语言模型。该软件帮助用户查询并整合隆格雷罗所说的“海量数据”，让操作员能够以普通英语与战场情报进行互动。
尽管美国国防部长皮特·赫格斯瑟近几个月公开与该公司发生冲突，将其称为“对国家安全的供应链风险”，但此次演习显示Anthropic仍发挥着重要作用。
Anthropic此前因推动明确禁止军方使用其强大的Claude AI模型对美国民众进行大规模监控，或为完全自主武器提供动力的防护措施，而激怒了特朗普政府官员。
需解决“道德伦理问题”，但“技术已成熟”
在摩洛哥沙漠中，一名美国士兵对让自主系统做出关键决策的想法向CBS新闻表达了疑虑。
“我们绝不能将决策责任委托给计算机，”这位要求不透露姓名的士兵表示，“计算机目前是我们的辅助工具，未来也将如此，但让我作为军官把手中的决策权交出去，我永远不会安心。”
“这是我们必须持续测试的战力倍增器，绝非万能解决方案，”他补充道。
4月30日，赫格斯瑟在参议院军事委员会表示，人工智能不会在战场做出致命决策，尽管他没有直接回应是否永远会如此。
“我们必须思考清楚道德和伦理问题，”负责美国非洲司令部的达格文·R·M·安德森将军对CBS新闻表示，他补充道：“技术已经成熟，不会消失，忽视它我们将自担风险。”
美国非洲司令部司令达格文·R·M·安德森于2026年5月，在摩洛哥阿加迪尔接受CBS新闻记者克里斯·利夫西采访，当时正值“非洲狮”联合军事演习期间。CBS新闻
“技术发展如此之快，我们正努力跟上步伐，”他承认，将致命责任从人类手中交给人工智能的想法“在我看来令人毛骨悚然，令人不安”。
“但不采用这项技术也很愚蠢，因为我们的对手会这么做，”他说。
“如果你选择不采用，我们将处于劣势，”他强调，“我绝不会让我们国家陷入这种境地。”
抢占机器人优势：“这关乎拯救生命”
摩洛哥演习中展示的另一项战场人工智能应用，是让士兵完全远离前线——用机器人取而代之。在这一领域最引人注目的私营国防承包商之一是总部位于西雅图的初创公司Overland AI，该公司在沙漠中对其ULTRA全自动车辆进行了性能测试。
操作员只需用笔记本电脑点击几下，就能告知ULTRA目的地，它将自主规划路线，避开危险和障碍物——并搭载机枪、地雷和炸药，以应对可能出现的任何情况。
Overland AI业务发展总监蒂姆·毕晓普对CBS新闻表示，ULTRA可通过向敌人实施掩护火力来保护士兵。它还可以布设地雷阻止敌军推进，并部署炸药突破敌军防线或防御工事。
ULTRA身高五英尺，配备坚固的越野轮胎，在演习中率先冲向交火线，其摄像头和传感器让远程操作员实时掌握其动向。
美国士兵在摩洛哥参与“非洲狮2026”军事演习， alongside 西雅图Overland AI公司制造的ULTRA全自动车辆。CBS新闻
目前，搭载的机枪仍由人类操作员远程操控，但毕晓普表示，未来实现该功能的自动化在技术上是可行的，届时机器将自行决定何时开火。
第一中尉文森特·加斯帕里对CBS新闻表示，他和战友们当天演练的突破作战是最危险的军事行动之一，而用机器人取代人类“无疑”将拯救生命。
“你不必再过多担心防护和生存问题，可以更快地移动，同时保护士兵，”他补充道，据他估计，在某次演习中，他们仅用两台机器人就取代了约40名士兵。
加斯帕里是第173空降旅刺刀创新团队的负责人，该团队是美国陆军创新的核心推动力之一，参与了此次演习。他承认对自主武器系统如何融入战争存在一些担忧，但表示他的动力源于保护战友的目标。
“我选择审视当下的工作，将其作为未来拯救生命数量的衡量标准，”他对CBS新闻表示，“我们必须抓住每一项优势，提升效率，成为最快、最强的一方，以比对手更快的速度做出决策，因为这关乎拯救生命。”
AI warfare is here, and CBS News got a look at the U.S. military training to use it on the battlefield
May 29, 2026 / 9:52 AM EDT / CBS News
Tan Tan, Morocco— In arid southern Morocco, where the Sahara Desert meets the Atlantic coast, the silence of the desert was shattered this month by the boom of explosions and crackling gunfire.
Plumes of smoke from conventional artillery filled the air as American forces took part in the multi-national African Lion 2026 military exercise. But the U.S. Army also used the war games to test an array of systems powered by artificial intelligence.
African Lion was the largest U.S.-led military exercise in Africa, carried out along with 30 partner nations to rehearse for the future of warfare. Increasingly, that future belongs to AI.
Alongside the military forces, more than a dozen private defense contractors showcased products and got feedback directly from soldiers as they vie for roles — and contracts — to help modernize the U.S. military.
Shortening the “kill chain”
As the soldiers practiced traditional battlefield tactics, a robot rolled silently across the Moroccan desert with a machine gun mounted on its roof. Drones lifted into the sky nearby carrying explosives, and another prototype quadcopter carried a nine-millimeter rifle.
One of the main applications of AI on display during the exercise was an effort to shorten the “kill chain” — the series of actions required to use lethal force, from the identification of a target to the moment a trigger is pulled.
U.S. Army Lt. Col. Ramon Leonguerrero told CBS News that personnel in the Joint Operations Center in Agadir, hundreds of miles from the mock battlefield, used an AI-driven platform made by American defense tech firm Palantir “to provide that rapid decision-making cycle faster than normal.”
U.S. military personnel work in the Joint Operations Center in Agadir, Morocco, during the joint African Lion 2026 exercise, which used an AI-driven platform made by American defense tech firm Palantir. CBS News
“Five years ago, this might have taken two or three hours,” he told CBS News of a decision made during one exercise. “We did it in three minutes.”
In that drill, there was a human at the end of the kill chain who approved the target and ordered an artillery unit to strike. But Leonguerrero told CBS News that autonomous systems that will decide when to pull the trigger without a human in the loop, to save more time, already exist. He would not say which, if any, real-world operations have used such systems.
At the operations center, dozens of people sat in front of a large screen and coordinated the movements on the ground.
The system powering much of the operation was Project Maven, the Pentagon’s flagship artificial intelligence initiative, created by Palantir. Maven ingests massive quantities of battlefield data and uses AI to identify patterns and prioritize information for commanders, such as determining what to target.
Anthropic’s Claude AI still vital despite row with Pentagon
According to multiple military and industry sources familiar with the systems used in the exercise, Maven’s interface with human operators relies on Anthropic’s Claude large language model. The software helps users query and synthesize what Leonguerrero called “an ocean of data,” allowing operators to interact with battlefield intelligence in plain English.
The drill showed Anthropic still plays a significant role, despite Defense Secretary Pete Hegseth publicly clashing with the company in recent months and labeling it a “supply-chain risk to national security.”
Anthropic has irritated Trump administration officials by pushing for guardrails that would explicitly prevent the military from using its powerful Claude AI model to conduct mass surveillance on Americans — or to power fully autonomous weapons.
“Moral and ethical issues” to figure out, but “the technology is there”
In the Moroccan desert, one U.S. soldier voiced skepticism to CBS News over the idea of letting an autonomous system make critical decisions.
“We can never delegate the responsibility of the decisions over to a computer,” said the soldier, who asked not to be named. “Computers enable us currently, and it’s my projection for the future, but I would never be comfortable delegating the decision that I hold as an officer.”
“It’s a force multiplier that we have to continue to test on, and it is not in any way a one-stop solution,” he added.
On April 30, Hegseth told the Senate Armed Services Committee that AI would not make lethal battlefield decisions, although he wouldn’t directly answer questions about whether that would always be the case.
“There are moral and ethical issues that we have to think through,” General Dagvin R.M. Anderson, who oversees the U.S. military’s Africa Command, told CBS News, adding: “The technology is there, it’s not going to go away, and we ignore it at our own peril.”
Gen. Dagvin R.M. Anderson, Commander of U.S. Africa Command, speaks with CBS News’ Chris Livesay in Agadir, Morocco, in May 2026, amid the African Lion joint military exercise. CBS News
“The technology’s emerging so quickly that we are working to keep up,” he acknowledged, calling the notion of AI taking over lethal responsibilities from humans “ghoulish to me, and it is disturbing.”
“But it’s also foolish not to adopt it because our adversaries will,” he said.
“If you choose not to adopt it, we will be at a disadvantage,” he stressed. “I would not be willing to put our nation into that position.”
Seizing a robotic advantage: “It’s about saving lives”
Another application for battlefield AI showcased during the exercise in Morocco involved keeping soldiers away from the front lines entirely — by replacing them with robots. One of the most visible private defense contractors working on that front was Overland AI, a Seattle-based startup that put its ULTRA fully autonomous vehicle through its paces in the desert.
Using a laptop, a remote operator can tell the ULTRA where to go with just a few clicks, and it will autonomously find its way there, avoiding hazards and obstacles — and packing a machine gun, mines and explosive charges to help deal with any that might pop up.
Overland AI’s Director of Business Development Tim Bishop told CBS News the ULTRA can be used to protect soldiers by deploying covering fire at an enemy. It can also lay mines to prevent enemies from advancing, and deploy explosives to breach an enemy line or structure.
At five feet tall with rugged off-road tires, the ULTRA sped ahead of soldiers toward the line of fire during the exercise, with its cameras and sensors keeping the remote operator aware of its movements.
U.S. soldiers take part in the African Lion 2026 military exercise in Morocco, along with an ULTRA fully autonomous vehicle made by Seattle-based Overland AI. CBS News
For now, the mounted machine gun is operated remotely by the human operator, but Bishop said it would be technically possible to automate that function in the future, with the machine deciding when to open fire.
1st Lt. Vincent Gasparri told CBS News that breaching operations like the ones he and his fellow soldiers practiced that day were among the most dangerous military operations, and replacing humans with robots would “undoubtedly” save lives.
“You don’t have to worry as much about protection and survivability. You can move faster and protect your soldiers while you do it,” he added, estimating that in one particular exercise, they had been able to replace about 40 humans with just two robots.
Gasparri leads the 173rd Airborne Brigade’s Bayonet Innovation Team, a driving force for innovation in the U.S. Army that took part in the exercise. He admitted to some apprehension over how autonomous weapons systems will be incorporated in warfare, but said he was driven by the goal of protecting his fellow soldiers.
“I choose to look around and measure the work we do today as a metric for the number of lives we’ll save in the future,” he told CBS News. “We have to take every advantage, find efficiency, be the fastest, the strongest, enable decision-making faster than the adversary, because it’s about saving lives.”