Claude AI used for missile, influence projects in UAE, Iran, Yemen: Anthropic

Military and government actors across the Middle East — from Iran to Yemen to the United Arab Emirates — have used Anthropic's Claude AI model for war planning, including missile projects, and for covert influence campaigns, the US technology company said in an intelligence report released Thursday.

The report covers malicious activity that Anthropic said it disrupted between December 2025 and August 2026 across several areas, including cyber operations, propaganda campaigns, surveillance and warfare.

Yemen

Among the report's revelations was that actors in Northern Yemen used Claude to develop guidance, navigation and control software that steers and stabilizes ballistic missiles.

Although Anthropic did not explicitly name the Houthis, who control northwest Yemen, it said it “identified a cell of threat actors based in northern Yemen” that was running three weapons development programs. 

These were a guided rocket that used a commodity phone-class flight computer with final-phase homing guidance; a multi-stage ballistic missile with a stated range goal above 2,000 kilometers; and a multivariant missile that included a hypersonic glide vehicle variant.

Anthropic gave an example of how the Yemen-based cell used Claude to try to guide missiles. They used Claude to integrate an open-source autopilot onto a phone-class flight computer, write the control and position-estimation software, edit the control settings, run a firmware build pipeline and perform a flight simulation.

The threat actors managed several Claude instances at once, assigning each a role, in what Anthropic described as “akin to how a lead would delegate work on a small engineering team." One instance wrote the code, another did research and a third reviewed the code the first instance produced.

Anthropic said its safeguards system blocked many of their requests, but not all, and the militants used a variety of tactics to evade the safeguards, including hiding their goals and the products the software was meant for. They also split their work across multiple sessions with Claude, so no single session revealed their full intent.

“These actors carried out a sustained effort to develop guided weapons, including using Claude to design guidance software. We do not have evidence the actors succeeded in fielding an operational device, but they did test-fire a guided rocket. This field test appears to have failed: within hours, the actors returned to Claude to work out why it failed,” Anthropic said in the report, adding that it identified the activity as part of its internal investigations into suspected weapons development.

Anthropic said it banned the suspicious accounts and shared threat information with public- and private-sector partners to mitigate risks.

Iran

Anthropic’s report also identified more examples of its software being used maliciously in the Middle East. The company said an Iran-linked actor used Claude to compile targeting handbooks on US Navy warships operating in the region. The entity used the AI model to collect and analyze publicly accessible data on US naval forces, including a roster of US personnel scraped from captions on public military photographs, ship and aircraft transponder identifiers, as well as commercial satellite-imagery query scripts. They also analyzed an inventory of public websites that exposed US naval movements, the report said. Anthropic said it banned the account carrying out the activity and shared threat intelligence with government authorities.

Separately, Iranian-state-aligned and Iran-linked accounts used Claude in separate operations involving propaganda, targeting US naval forces, domestic surveillance and profiling people in Israel and in the Jewish diaspora.

Anthropic said it identified and removed three Iranian accounts linked to government propaganda institutions: The institutions included the Islamic Culture and Communication Organization under the Ministry of Culture and Islamic Guidance, the Islamic Propaganda Office of Khorasan Razavi, running a cognitive warfare command room out of a Mashhad seminary distributing content aligned with Islamic Revolutionary Guard Corps narratives and the Islamic Propaganda Organization’s Bina Cultural Observatory.

Claude was used by the Iranian accounts to build campaign plans, doctrine manuals, persona systems, target databases and ministerial planning documentation, Anthropic said.

Al-Monitor contacted the Iranian foreign ministry for comment.

United Arab Emirates

The report also said UAE government officials allegedly used Claude to run an influence operation against the Muslim Brotherhood and counter criticism of the UAE’s role in the Sudan conflict, where Abu Dhabi has been accused of backing the Rapid Support Forces that are fighting the Sudanese military. Since 2014, the UAE has designated the Brotherhood as a terrorist organization.

The operation used around 300 fake social media accounts, created a front nongovernmental organization, ghostwrote testimony to be used at the UN and compiled files on European politicians, journalists and UN officials.

Al-Monitor contacted the UAE Foreign Ministry for comment.

Why it matters: The findings provide a rare documented example of a commercially available AI model being incorporated into an active weapons-development program. Anthropic did not identify the northern Yemen cell as belonging to the Houthis, and said it had no evidence that the actors successfully fielded an operational weapon. But the group’s location, the sophistication of its missile projects and its reported test launch are likely to fuel scrutiny of possible Houthi use of advanced AI tools.

The Iran- and UAE-linked cases point to a broader challenge. According to Anthropic, regional actors used Claude not only for technical military work but also to organize propaganda, surveillance and influence campaigns at scale.

The activity unfolded against the backdrop of the US-Israel-Iran war, which began on Feb. 28 and overlapped with much of the period covered by Anthropic’s report. 

Know more: Before the war began, the US Pentagon had already deeply integrated Claude into its defense infrastructure, mainly by embedding it with Palantir’s Maven Smart System. Anthropic’s AI software has had many military applications. For example, Claude has acted as a decision support system, helping military planners sort vast amounts of satellite data to strike 1,000 targets during Operation Epic Fury, the original US and Israeli air strikes on Iran that began the war. Claude was also used in the January operation that led to the capture of Venezuelan President Nicolas Maduro.