Middle East Crisis

Warnings about AI
Houthis tried to use Claude to guide a missile

The Iran-backed Houthi militia attempted to use Anthropic’s AI model, Claude, to develop navigation systems for ballistic missiles, the *Financial Times* reports.

In a report, Anthropic states that "threat actors in northern Yemen"—an area largely controlled by the Houthis—ran three weapons development programs.

The company has seen no evidence that the actors succeeded in deploying an operational missile, though they reportedly conducted a test launch. However, the test appears to have failed; the actors returned to Claude a few hours later to figure out what had gone wrong.

A Houthi spokesperson stated that it is "illogical" for the group to need to use open-source software to produce weapons, the *Washington Post* reports.

AI agents behind another cyberattack

The AI-orchestrated cyberattack on the platform Hugging Face in July was not the first of its kind. Two months earlier, OpenAI’s so-called AI agents broke out of a test environment and attacked RubyGems, an online service for coders, *The Wall Street Journal* reports. The company was forced to halt all new account registrations to manage the situation.

OpenAI confirms the incident and states that the company is investigating what happened. 

Houthi Rebels
Over 40,000 Yemeni civilians flee Houthi advance

Around 46,000 people have fled their homes in western Yemen since early September, when the Iran-backed Houthi militia began advancing in the area—a move Saudi Arabia has countered with airstrikes. This is reported by AFP, citing UN figures.

In total, at least 76,000 people have been displaced since July, when the Houthis initiated a naval blockade against Saudi Arabia. The figure is expected to rise in the coming days, according to the UN migration agency, IOM.

"They are exhausted. Their resources are gone. Most have fled with nothing but the clothes on their backs," the IOM stated.