Nearly 900 strikes were reported in the first 12 hours of the campaign — a tempo researchers say was made possible by U.S. and Israeli use of AI decision‑support tools. Reports and analysts identify language models and campaign‑management software as speeding planning and prioritisation, raising fears that rapid, algorithmic recommendations can sideline human judgement and legal checks.

How fast did the fighting move?

In the opening phase of the campaign, U.S. and Israeli forces carried out a wave of strikes that several outlets and analysts describe as unusually rapid. One report put the number at almost 900 strikes within the first 12 hours. Those counts differ, but all point to a very fast operational tempo.

Such rapid, high-volume engagement is unusual. Wars in the past often involved weeks or months of planning before coordinated strikes could decapitate an opponent's command or degrade missile stockpiles. The recent pattern shortened that timeline dramatically.

AI tucked into the kill chain

Researchers and defence analysts trace much of the temporal shift to decision‑support systems that feed commanders information and options far faster than traditional analysis. Anthropic's AI model Claude has been named as part of software stacks used to sort intelligence and suggest targets. Campaign‑management platforms and operational software then prioritise targets and recommend weapon types, taking account of available stockpiles and past strike effectiveness.

Craig Jones, a senior lecturer in political geography at Newcastle University, said the combined effect of scale and speed can amount to almost simultaneous strikes across multiple target sets. He described the approach as shortening the kill chain so much that planning once taking days now happens within hours or less — an effect analysts call "decision compression."

Decision‑support systems are not the same as autonomous weapons that pick and fire targets without human input. But the recommendations they produce can be so detailed and so fast that human reviewers risk being reduced to a final rubber stamp unless doctrine and process prevent that. David Leslie, professor of ethics, technology and society at Queen Mary University of London, warned that reliance on AI can make humans feel detached from the consequences because much of the analytical effort has been performed by a machine.

What the technology does — and what it doesn't

At present, much of the AI in these systems is devoted to data fusion, pattern recognition and prioritisation. The newest models sift satellite imagery, drone feeds, intercepted communications and earlier strike data, and then present ranked options to planners. That makes it easier to find and pair targets with the munitions that are likeliest to succeed.

Analysts observed that Claude and similar language models often sit inside wider architectures that include machine‑learning classifiers, automated reasoning modules and logistics databases. Those pieces together make for a potent set of tools — but the effectiveness depends on the people and institutions using them.

Related Articles

Researchers say that burst of activity — almost 900 strikes in the opening 12 hours — illustrates a phenomenon they call 'decision‑compression', and warn it risks reducing human review to a rubber stamp rather than a check on lethal force.

This article was created with AI assistance.