AI Enters War, Pentagon Used Claude in Iran Attack

Artificial intelligence is no longer something the world talks about in theory when it comes to war. It is already in use. Recent reports show that the United States military relied on several AI tools, including Claude from Anthropic, during its strikes on Iran in late February.

While attention has focused heavily on Claude, the reality is broader. The Pentagon has spent years quietly building an AI ecosystem that now supports intelligence work, target identification and battle simulations. What happened during the Iran strikes did not start overnight; it was the result of long planning and steady integration.

The US Department of Defence has, over time, signed agreements with multiple major technology companies. These include OpenAI, Google and xAI. Each of these companies provides different AI tools that help the military process information and plan operations.

In July 2025, the Pentagon awarded contracts worth up to $200 million through its Chief Digital and Artificial Intelligence Office. The goal was clear. The military wanted advanced AI systems that could work across national security, intelligence gathering and war planning. These were not one-time purchases but ongoing usage agreements that allowed gradual deployment.

Anthropic became deeply involved in US defence systems from late 2024. In November of that year, it partnered with Palantir and Amazon Web Services to supply Claude to defence and intelligence agencies, including classified environments. By June 2025, the company had rolled out Claude Gov, a version built specifically for government and national security tasks.

By late 2025, Claude Gov was already in active use across several US intelligence and defence agencies. Public records later showed that only a small portion of the contract funds had been paid by early 2026, pointing to steady integration rather than sudden adoption.

During the Iran strikes, reports say the military relied mainly on AI models from the Claude family. These tools were used to analyse intelligence, identify potential targets and simulate different battle outcomes. The AI systems helped process large volumes of data, including satellite images, intercepted communications and signals intelligence.

For intelligence assessment, Claude reviewed information to help evaluate threats and enemy positions. In target identification, the system assisted in locating and cross-checking high-value targets, including leadership compounds and strategic facilities. For simulations, the AI helped model possible strike outcomes and assess risks before decisions were made.

Despite its deep involvement, the reports make one thing clear. Claude did not act as a weapon. It did not independently make lethal decisions or control any weapons systems. Its role remained that of a support tool, offering summaries, insights and simulations to human decision-makers.

This distinction became especially important when a dispute erupted between Anthropic and the administration of Donald Trump. Just days before the Iran strikes, Trump issued a directive banning federal agencies from using Anthropic’s tools following disagreements over usage rights and ethical safeguards.

Defence Secretary Pete Hegseth was involved in designating Anthropic as a supply chain risk. Agencies were told to stop using its technology, triggering a phase-out period expected to last several months. Even so, Claude remained in use during the Iran operation because its integration into intelligence and planning systems was already deep.

At the heart of the dispute was how far the Pentagon could go with AI. The military wanted broader rights to use Claude “for all lawful purposes”, including possible battlefield targeting and weapon support. Anthropic refused to remove its ethical constraints, which are part of its Constitutional AI framework.

Claude is programmed to reject use in fully autonomous lethal weapons. It also refuses to approve strikes without human oversight and will not support mass domestic surveillance of US citizens. These limits became a breaking point in negotiations.

As Anthropic stepped back, other companies moved in quickly. OpenAI reached new agreements to deploy its models on classified networks, while Google expanded its Gemini for Government tools for unclassified use. Elon Musk’s xAI also secured deals to supply Grok for Government, including classified deployments.

Platforms such as GenAI.mil, ChatGPT, Gemini and Grok are now used widely for unclassified military tasks. Classified systems, once dominated by Anthropic’s deployments, have since expanded to include OpenAI and xAI tools.

What this shows is simple and unsettling. By early 2026, AI has become deeply embedded in US military planning and execution. The Iran strikes only made that reality visible to the world.

For observers, including those following events from Nigeria and across Africa, the lesson is clear. Warfare has entered a new phase. Decisions may still rest with humans, but machines are now firmly in the room. As The Gazette News (Nigeria) continues to monitor global security trends, one thing is certain: the era of AI in warfare is no longer coming. It is already here.