3,000 strikes, zero oversight: AI’s quiet war in Iran

The U.S. military’s 3,000-target strike campaign in Iran wasn’t just supported by AI—it was run by it. Generative models handled intelligence synthesis, target prioritization, and logistical routing, according to The Wall Street Journal’s confirmation of earlier reports. This isn’t a demo or a PowerPoint slide about ‘future warfare’; it’s deployed, live-fire integration with a casualty-ledger attached.

The real shift here isn’t the use of AI—militaries have leaned on algorithms for years—but the scale and autonomy. Previous systems flagged potential targets for human review; now, generative models are ‘co-piloting’ decisions in near-real time. That’s a leap from ‘tool’ to ‘partner,’ and one made without proportional investment in oversight. The Decoder’s original report noted the ‘underinvested’ audit trails—a polite way to say the guardrails are MIA.

This isn’t just about Iran. It’s a template. The same stack—Palantir’s Maven Smart System, Scale AI’s data pipelines, and custom DoD models—is being pitched to NATO allies as the new standard. The sales pitch writes itself: faster decisions, fewer boots on the ground, ‘scalable deterrence.’ The fine print—who audits the auditor?—gets a footnote.

The hype filter here needs to separate two things: AI as force multiplier (real) and AI as accountable decision-maker (fiction). Military contractors like Anduril and Scale AI will tout this as proof their tech ‘works in the wild.’ Technically true. But ‘works’ for whom? The 2023 RAND report on AI in warfare warned that speed and accuracy often trade off against explainability—a gap this campaign seems to embrace, not close.

Developer signals are mixed. GitHub repos for military-adjacent AI tools show upticks in forks of open-source targeting frameworks, but the chatter skews pragmatic: ‘If the DoD won’t audit, we’ll reverse-engineer the outputs.’ Meanwhile, Big Tech’s ethical AI teams—the ones who used to flag risks like this—are being quietly dismantled.

The competitive advantage here isn’t just America’s. It’s for any state willing to deploy AI before the oversight catches up. China’s ‘AI +’ military doctrine is watching closely—not for the tech, but for the permission structure. If the U.S. normalizes unaudited strikes, the barrier to entry drops for everyone.