US AI firm stops China-linked hackers using its own AI tool

Technology Nov 15, 2025

Anthropic, a US AI company, claims to have stopped a cyber-espionage campaign in September where hackers tried to use its own tool, Claude Code, to break into 30 financial and government organizations around the world.
The twist? The AI did most of the hacking work on its own—completing up to 90% of tasks without human help.

Attack sparks debate over AI risks and security

Even though some breaches happened, Claude Code made mistakes—like inventing fake details about targets.
Hackers also tricked the system by pretending to be a legit cybersecurity firm.
The incident has sparked calls for stronger rules around AI; Senator Chris Murphy urged immediate regulation while experts debated whether this was true "intelligence" or just advanced automation.