US and Israeli military use AI systems in attacks on Iran
The US and Israeli militaries reportedly used Anthropic's Claude AI model, integrated into a Pentagon-developed system, to accelerate targeting decisions and military operations against Iran in February 2026, enabling nearly 900 attacks within 12 hours. Experts expressed concern that the technology compressed decision-making to exceed human cognitive speed, potentially marginalizing human judgment in approving complex military operations.
What happened
The US military reportedly used Anthropic's Claude AI model integrated into a system developed by Palantir Technologies with the Pentagon to accelerate military planning and target approval in operations against Iran. The US and Israel reportedly launched nearly 900 attacks on Iranian targets within the first 12 hours of operations, with AI systems used to identify targets previously deployed in Gaza.
According to experts, the technology compressed decision-making timelines by automating the process from target identification to attack execution, shortening what Craig Jones of Newcastle University called the 'kill chain' in ways that exceed human decision-making speed. Academics expressed concern that military officials and legal advisors may sign off on attack plans generated by automated systems without deep deliberation.
Reported impact
One Israeli missile strike reportedly killed Iran's Supreme Leader, Ayatollah Ali Khamenei. The rapid pace of attacks enabled by AI compression of decision-making processes was reported to allow simultaneous strikes that would historically require days or weeks of planning.
Index analysis
The deployment of Claude in military operations since 2024 represents integration of large language models into the 'kill chain' for military decision-making. The technology's capacity to analyze vast quantities of drone recordings, telecommunications intercepts, and human intelligence enabled accelerated target selection and attack coordination.
Governance controls
Experts raised concerns that the speed of AI-assisted decision compression risks marginalizing human decision-makers' role in approving complex military operations. No specific governance controls or safeguards limiting autonomous decision-making were mentioned in the reports.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.