Anthropic, developer of AI software including Claude (seen at left). ©2026 Anthropic
One section of the report, which was released on Thursday, described a “cell of threat actors based in northern Yemen” trying to use the company’s Claude AI to create a ballistic missile guidance system.
“The actors used Claude Code in place of human software engineers to develop the guidance, navigation and control software that steers and stabilizes a flying vehicle,” the company wrote. “For example, they used Claude to integrate an open-source autopilot onto a phone-class flight computer, writing the control and position estimation software, tuning the control settings, running a firmware build pipeline and performing a flight simulation.”
Using large language models like Claude to write functional software code without knowing how to write a programming language is commonly known as “vibecoding” and has been one of the revolutionary developments driving Anthropic’s rise to what is likely to be the largest initial public offering in history later this year.
The report did not name the Yemeni “threat actors,” but its description of their development of “a multi-stage ballistic missile with a stated range goal above 2,000 kilometers” likely refers to the Iran-backed Houthis, who have have used ballistic missiles to attack Israel, Saudi Arabia and other targets around the region.
Anthropic said that it “disrupted” the Yemeni cell and informed “public- and private-sector partners to mitigate risks.” But the group used Claude to make progress in its weapons development programs, Anthropic said.
“We do not have evidence the actors succeeded in fielding an operational device, but they did test-fire a guided rocket,” the company said. “This field test appears to have failed: within hours, the actors returned to Claude to work out why it failed.”
The company says it banned the accounts, but “we have evidence that the actors had already built an offline simulation toolkit that does not rely on Claude or other engineering computing environments” to carry out the work.
In another case study, Anthropic said that “an Iran-nexus threat actor” used Claude to collect open-source data to generate targeting recommendations against U.S. Navy facilities in the Middle East.
“The compiled material included a roster of U.S. personnel scraped from captions on public military photographs, publicly accessible ship and aircraft transponder identifiers, commercial satellite-imagery query scripts and an inventory of public websites that exposed U.S. naval movements,” the company said.
The 154-page security report from the AI company also warned of suspected attempts to carry out biological weapons development and to use its models to support a propaganda campaign from Iran’s Ministry of Culture and Islamic Guidance to wage “explanatory jihad.”
“We’re publishing this work, because we believe we have a responsibility to disclose malicious misuse of our services,” the company said. “As models become increasingly capable, their risks will increase, unless AI developers and society’s defenders act to make them safer.”
