ISTANBUL

To carry out strikes on 1,000 targets on the first day of its assault on Iran, the US military relied on the most sophisticated artificial intelligence system it has deployed in combat to date, technology that the Pentagon may find difficult to abandon even as it ends its relationship with the company behind it.

The tool, known as the Maven Smart System and developed by data analytics company Palantir Technologies, analyzes massive volumes of classified intelligence from satellites, surveillance platforms, and other sources.

The system produces rapid insights that help commanders identify and prioritize targets in real time during operations involving Iran, the Washington Post reported, citing three individuals familiar with the program.

Integrated into the platform is Claude, an AI model created by Anthropic.

The technology, however, was barred from use by the Pentagon last week due to tense negotiations over how it could be used during wartime.

Despite that dispute, military planners say the pairing of Claude with the Maven system has evolved over the past year into a widely used operational tool.

According to two sources familiar with the program quoted by the newspaper, it is now used on a daily basis throughout much of the US military.

As preparations for a potential strike on Iran were being developed, the Maven Smart System, enhanced with Anthropic’s Claude AI model, reportedly generated hundreds of possible targets, provided exact geographic coordinates, and ranked them based on operational importance, according to two sources familiar with the process.

The integration of Maven and Claude has accelerated the pace of military planning, allowing commanders to move from weeks of battlefield preparation to near-real-time operational decisions, one of the sources told the Washington Post.

The system also analyzes strike results after they occur, which helps in determining the operation's effectiveness.

US authorities have previously used Claude in counterterrorism operations and other offensives, including the capture of Venezuelan President Nicolas Maduro.

However, according to two sources, the newspaper states that the campaign involving Iran marks the first time the technology has been deployed during a large-scale war operation.

Despite the AI system’s role in supporting US military activities, relations between Dario Amodei, CEO of Anthropic, and the administration of Donald Trump have deteriorated.

Just hours before the Iran bombing campaign began, Trump announced that US government agencies would be barred from further using Anthropic’s technology, giving them six months to remove it from their systems.

The decision followed a dispute between the company and the Pentagon over how the tools could be used, particularly on mass domestic surveillance and fully autonomous weapons.

The military will continue to rely on the system temporarily until alternative technologies are introduced, unnamed sources cited by the Washington Post as saying.

One person familiar with the matter added that the Pentagon has become heavily dependent on the AI platform, and if Anthropic attempted to cut off access, the administration could invoke government authorities to retain the technology until a replacement is ready.

The Pentagon did not immediately respond to requests for comment, according to the daily.

Anthropic spokesperson Eduardo Maia Silva and Palantir representative Lisa Gordon also declined to comment on the matter. The Wall Street Journal was the first to report on the use of the Claude AI system in attacks against Iran.

The deployment of advanced generative artificial intelligence in the attacks on Iran comes amid an intensifying debate over both the ethical implications and the rapid integration of AI technologies into modern warfare.