Iran-linked operatives and Houthi rebels used Anthropic’s Claude AI to conduct military reconnaissance, develop weapons and build surveillance tools targeting opponents, according to a September 2026 threat report from the company.

AI turned against the US Navy

Anthropic said an Iran-linked actor used Claude to analyse public ship and aircraft identifiers, commercial satellite imagery and information on US naval movements in the Middle East. The operation also extracted names of US military personnel from public photographs and researched vulnerabilities in communications systems used aboard warships.

The actor examined weaknesses affecting Cobham Sailor VSAT terminals, Cisco equipment and Schneider Electric EcoStruxure systems. Anthropic banned the account, added new detection measures and shared its findings with government authorities.

Claude became a weapons engineer

In northern Yemen, a Houthi-controlled cell allegedly used Claude Code across three weapons programmes, including a guided rocket, a multistage ballistic missile designed for ranges above 2,000km and an R2000 missile family featuring a hypersonic glide vehicle variant.

The AI was reportedly used to write guidance, navigation and control software, integrate an open-source autopilot with a phone-class flight computer, tune parameters, build firmware and conduct flight simulations. Multiple Claude instances effectively substituted for software engineers, researchers and code reviewers.

Iran also weaponised AI for surveillance

The technology was not limited to battlefield applications. One Iran-linked security unit used Claude to analyse 155,216 tweets and profile 6,388 opposition figures in a single year.

Another group reportedly used the model to develop domestic tracking technology, including the production-deployed “al-Najm al-thāqib” Firefox extension, designed to harvest identities across major social platforms.

Anthropic faces a wider threat

Anthropic said it has banned 16 Claude accounts linked to Iranian paramilitary and domestic security agencies, but warned that enforcement does not necessarily identify every account being used.

Iran and the Houthis are also not alone. Anthropic said Chinese and Russian actors are actively using Claude in military programmes, highlighting a growing problem for AI companies: technology built in America is increasingly being repurposed to improve the capabilities of its strategic rivals.

Iran and Houthi rebels used Anthropic's Claude AI to target US warships, build hypersonic missiles

I feel a bombing campaign may be coming soon. Guess where?

Target folders being studied now at CENTCOM.

Love a good shock and awe video.