A kingdom under pressure from a movement it cannot easily dislodge has turned to its most powerful ally for help, and the help offered is not the strike it asked for but a faster way of seeing and striking. Between one hundred and two hundred American personnel now stand beside Saudi officers in a joint command, feeding them a system that sorts satellite images and drone footage into lists of targets. Such tools have already shortened the gap between seeing a threat and acting on it, a gap that in past wars gave time for doubt. Whether speed in Yemen will bring the mountain-hidden Houthis to h
Saudi Arabia Turns to US AI Targeting in Yemen, Raising Questions on Effectiveness and Civilian Harm
A Google Earth for war, now lent to Riyadh
Why does a software tool matter so much in this Yemen story?
Because it changes speed. In Iran, MSS-assisted forces hit over 1,000 targets in a day, against about 100 before. Saudi Arabia is getting something similar.
And the US isn't fighting directly?
Media reports say Trump refused that request from Mohammed bin Salman. Intelligence and targeting help is the route instead.
Worth noting the whole Saudi arrangement rests on CNN quoting unnamed officials. The 100 to 200 personnel figure isn't officially confirmed, and the system is described as similar to MSS, not identical.
What are the limits of the technology?
The Houthis move fast with truck-mounted missiles and small drones in rough terrain. Dust, wind and decoys also degrade the system.
Those are noted weaknesses, but nobody has shown how it performs in Yemen. That's still unknown.
Then there's the school in Iran.
Almost 200 killed, mostly children aged seven to 12. A human is meant to verify targets, yet the source says that doesn't always happen.
The reporting ties MSS to targeting that site, but it doesn't lay out how the decision was made. And the Gaza civilian toll claims are reported, with analysts saying safeguards were loosened.
So can AI defeat the Houthis?
Loewenstein doubts it, and Chatham House's al-Muslimi says the Houthis are not deterred by destruction. Luke would say those are expert opinions, not results.
El Pulso
- Washington reportedly refused a Saudi request for direct military action against the Houthis, offering AI-driven targeting intelligence instead.
- Roughly 100 to 200 US personnel are embedded in a new joint command, helping Saudi officers use a tool modeled on the Maven Smart System.
- The system compresses the kill chain so sharply that targets can be struck faster than human judgment may fully weigh them.
- Earlier campaigns have not weakened the Houthis, who are mobile, ideologically unyielding and hard to find in hills and canyons.
- The school strike in Iran, which killed nearly 200 children and staff, and the Gaza record hang over the project as a warning of civilian harm.
A kingdom under pressure from a movement it cannot easily dislodge has turned to its most powerful ally for help, and the help offered is not the strike it asked for but a faster way of seeing and striking. Between one hundred and two hundred American personnel now stand beside Saudi officers in a joint command, feeding them a system that sorts satellite images and drone footage into lists of targets. Such tools have already shortened the gap between seeing a threat and acting on it, a gap that in past wars gave time for doubt. Whether speed in Yemen will bring the mountain-hidden Houthis to heel, or simply add to the long record of civilian suffering that follows such tools, remains an open question that the terrain and the people beneath it will answer.
Saudi Arabia has been asking its allies for help against the Houthi movement, which has pushed deeper into Yemen's coast and renewed attacks on the kingdom's airport and oil facilities. According to a CNN report citing unnamed American officials, Washington turned down a request for direct military action from Crown Prince Mohammed bin Salman. Instead, it is offering intelligence and an AI targeting tool, with between 100 and 200 US personnel in Saudi Arabia helping run a new joint command built around it.
The tool is similar to the Maven Smart System, which sifts live satellite and drone feeds, labels likely targets such as launchers and vehicle convoys, and suggests how to hit them. Its chief value is speed. In the first day of the American attack on Iran, it helped enable more than a thousand strikes, far above the roughly one hundred a day that had been typical before.
Experts doubt it will work well in Yemen. Farea al-Muslimi of Chatham House says the Houthis' ideology treats hardship and martyrdom as meaningful, so civilian losses do not deter them. Their mobility, from truck-mounted missiles and drone swarms fired from hills, means fighters may vanish before a strike lands, and the tool performs poorly in dust, dense terrain and against decoys.
The deeper worry is the human check. A person is meant to verify every flagged target, but that does not always happen. A missile strike on a primary school in southern Iran killed almost 200 people, most of them children, and Israel's Gaza experience shows similar AI systems can bring faster strikes at a heavy cost in civilian lives. What to watch now is whether Yemen follows that same path.
Saudi Arabia, under pressure from the Houthi movement in neighboring Yemen, has been asking its allies for help. According to media reports, one request was turned down: President Donald Trump declined when Crown Prince Mohammed bin Salman asked the United States to take direct military action against the Houthis. What Washington is providing instead is intelligence, including a targeting tool driven by artificial intelligence.
The detail comes from a September CNN report that quoted several unnamed American officials. It said between 100 and 200 US soldiers are in Saudi Arabia as part of a new joint force command set up to make this possible. Their job is to help Saudi officers use a tool that shows what is happening on the battlefield in real time. Officials described the system as similar to the Maven Smart System, or MSS, which the US military has used while looking for targets in Iran, Iraq and Syria.
MSS grew out of Project Maven, the Pentagon's flagship military AI program, launched in 2017. Analysts at the Center for Strategic and International Studies in Washington wrote in June that it gathers, organizes and visualizes huge streams of intelligence, surveillance and reconnaissance data. It finds patterns in material such as live satellite feeds and drone camera footage, then labels possible targets: rocket launchers, tanks, suspicious vehicle convoys. It also calculates the best way to hit them. Human analysts once had to comb through many separate sources, which the CSIS researchers described as nearly impossible with hours of footage. Tech commentator Bilawal Sidhu, a former Google product manager, has called it a "Google Earth for war."
The main effect is speed. MSS compresses the kill chain, the interval between detecting a target, judging whether it is dangerous and ordering a strike. When the US attacked Iran in late February, the system helped enable hits on more than 1,000 targets in the first 24 hours. American forces had previously struck about 100 targets a day.
The Yemen front had been relatively calm after an informal ceasefire in 2022. The Houthis have held large parts of northern and western Yemen since 2015, while the internationally recognized government and other groups, backed by Saudi Arabia and the United Arab Emirates, hold territory in the south and east. A new Houthi campaign beginning in September has extended the group's control over more of the coast and renewed attacks on Saudi Arabia, including strikes on its main airport and oil facilities. Earlier campaigns by Saudi Arabia, the US and others have not significantly weakened the group.
Farea al-Muslimi, a Yemen research fellow at the UK think tank Chatham House, told DW that part of the reason is the Houthis' zero-sum outlook. He said the group's founder held that modern Muslim societies had grown too soft and that hardship is a divine test. "They don't care if the Israelis bomb the airport or the US bombs a hospital," he said, adding that this serves them because their ideology centers on martyrdom and sacrifice.
There are technical limits as well. The Houthis are highly mobile, using truck-mounted ballistic missiles and small drone swarms from hills and canyons, so fighters may be gone before a strike lands, even with AI help. MSS is known to work best in sparsely populated areas, and shifting desert terrain, dust storms, high wind, snow, dense foliage and deliberately planted decoys can all reduce its success rate.
The ethical problem is harder. A human is supposed to verify every target the tool flags and decide how to respond, but that does not always happen. Among the sites MSS targeted on the first day of the Iran war was a school: an American missile killed almost 200 people at a primary school in southern Iran, most of them children aged seven to 12.
Israel's experience in Gaza offers another reference point. Its Gospel system, or Habsora, is the closest in function to MSS. Analysts Frederic Wehrey and Andrew Bonney wrote in War on the Rocks that AI improved commanders' awareness, the speed of strikes and troop survivability, but that those gains were overshadowed by civilian deaths reportedly tied to loosened or removed safeguards. Journalist Antony Loewenstein, author of "The Palestine Laboratory," told DW he sees little evidence that AI will change much on the ground "except for causing huge amounts of civilian casualties," and expects the US and Saudis to find the Houthis have "an incredibly high tolerance level for destruction." What to watch is whether a shorter kill chain in Yemen brings the same civilian toll seen in Gaza and Iran.
Citas Notables
I don't think there's much evidence at all that using an AI tool is going to make much difference on the ground, except for causing huge amounts of civilian casualties.— Antony Loewenstein, independent journalist, to DW
They don't care if the Israelis bomb the airport or the US bombs a hospital.— Farea al-Muslimi, Chatham House, to DW