A major artificial intelligence company said that Iranian bad actors used its publicly accessible software to target U.S. service members in the Middle East.
The Thursday Anthropic report covers the time period between December 2025 and August 2026 and highlights the misuse of its large language model Claude, which it says the threat actor used to make targeting recommendations against U.S. Navy forces in the region.
“The threat actor used Claude to compile targeting handbooks…to identify and track naval positions based on open-source information,” the report said.
The bad actor used the AI software to collect information that included a list of U.S. service members it had sourced from public military photographs, ship and aircraft transponders — which broadcast a specific code that identifies them — commercial satellite-imagery query scripts that help locate exact geographic visuals, and a breakdown of public websites that track U.S. Navy movements.
The Iranian threat also used Claude to learn more about weaknesses in U.S. Navy shipboard systems that could be exploited.
Anthropic said it banned the account after it learned what it was doing, created safeguards to prevent similar activity in the future and shared the information with the appropriate government agencies to address the problem.
The report detailed a litany of other misuses of its Claude program, including a cell of Yemeni bad actors that used the program to develop targeting software for guided missiles.
The concerns over inappropriate misconduct involving Anthropic have grown to include individuals within the company too.
A researcher named Jacob Coxon from Anthropic and OpenAI resigned Tuesday, calling into question the ethics of the company in a lengthy X post.
“Neither company is acting responsibly,” said Coxon. “They are racing straight to self-improving superintelligence and gambling with our lives.”
The U.S. military previously used Anthropic but severed ties with the AI platform after it refused to agree to Pentagon terms that it said would open up Claude’s use for mass surveillance or use in fully autonomous weapons.
Riley Ceder is a Navy reporter at Military Times, where he covers issues directly affecting service members and their families, as well as breaking news.
Read the full article here