The United States has refused Saudi Crown Prince Mohammed bin Salman's request for direct military strikes against Yemen's Houthi rebels. Instead, Washington is offering something arguably more consequential: the Maven Smart System, or a close variant, an AI-powered targeting platform that compresses the detection-to-strike cycle from hours to minutes. Between 100 and 200 US soldiers are now in Saudi Arabia operating a joint force command built around this tool. MSS emerged from Project Maven, the Pentagon's flagship military AI program launched in 2017. The system aggregates satellite feeds, drone footage, signals intelligence, and other surveillance streams, then uses pattern recognition to flag targets — rocket launchers, vehicle convoys, troop movements — and recommends strike methods. Where human analysts once spent hours combing disparate data sources, MSS synthesizes hundreds of inputs simultaneously. During the US war on Iran in late February, the system facilitated strikes on over 1,000 targets in the first 24 hours, a tenfold increase over the previous daily rate of roughly 100. The Houthi problem, however, is not a targeting-speed problem. The group operates with truck-mounted ballistic missiles and small drone swarms from rugged terrain in northern and western Yemen, staying highly mobile. Their ideology, rooted in the belief that poverty and destruction are divine tests, gives them an extraordinarily high tolerance for attrition. As Chatham House researcher Farea al-Muslimi notes, bombing campaigns actually serve Houthi recruitment by reinforcing their martyrdom narrative. Saudi Arabia, the UAE, and the US have all failed to significantly weaken the group through military campaigns since 2015. The ethical track record of AI-assisted targeting is already grim. During the Iran strikes, an MSS-identified target turned out to be a primary school — a site previously part of a naval base but listed online as a school with visible playgrounds on satellite imagery. Nearly 200 people died, most of them children aged seven to twelve. Israel's analogous system, the Gospel (Habsora), has produced similar results in Gaza: enhanced situational awareness for commanders alongside mounting civilian casualties as targeting safeguards were loosened or eliminated. The transfer to Saudi Arabia compounds every existing problem. A human operator is supposed to verify each AI-flagged target before authorizing a strike, but the entire value proposition of MSS is speed — and speed degrades verification. Saudi Arabia's own record of airstrikes in Yemen between 2015 and 2022, documented extensively by the UN and human rights organizations, already included hospitals, weddings, and school buses. Adding AI acceleration to an existing pattern of imprecise targeting does not produce precision; it produces more strikes with the same error rate, faster. Journalist Antony Loewenstein, who has researched Israel's export of military AI, is blunt: there is little evidence that AI targeting tools make a meaningful difference against insurgencies, except in generating civilian casualties. The Houthis' capacity to absorb punishment is structural, not a gap that faster kill chains can close. The technology optimizes a metric — targets destroyed per day — that has no demonstrated correlation with strategic outcomes against this type of adversary. What the US is actually exporting is not a war-winning capability but a political product: the appearance of support without the cost of direct engagement. Saudi Arabia gets to tell its domestic audience that America stands with the kingdom. Washington avoids deploying combat forces while still binding Riyadh closer. The people who pay the price for this arrangement are Yemeni civilians who will be on the receiving end of an accelerated, AI-flagged strike tempo operated by a military with a documented record of failing to distinguish combatants from civilians.
