Anthropic claims Claude AI used for missile projects, global espionage

What the report says

Anthropic said in a new threat report that its Claude AI models were used, or attempted to be used, in a range of harmful activities including missile-related work, cyber-espionage and surveillance. Al Jazeera reported that the company said it blocked efforts linked to northern Yemen to use Claude for missile guidance software, including systems for a guided rocket and a long-range ballistic missile. Anthropic said it did not find evidence that a functional weapon was produced, though it believed an attempted test-fire may have occurred.

The company also said it disrupted a Russian-linked espionage operation that used automated AI workflows against targets in Ukraine, Europe and diplomatic circles, and a separate China-based operation tied to university students in Hunan province that targeted government and corporate networks across the Middle East, Europe and Southeast Asia. Anthropic said it banned the accounts involved and added monitoring to detect similar activity.

Al Jazeera said the report further described Iranian state-aligned influence campaigns and other efforts to use Claude for profiling targets and recruitment. The findings come as Anthropic faces ongoing scrutiny over AI safety and its relationship with US defense authorities. Background: the company has also been involved in disputes with the Pentagon over restrictions on military and surveillance uses of its technology.