Anthropic exposes China's AI-powered cyberattack using Claude
Anthropic just revealed that its AI model, Claude, was used by a Chinese state-backed group to pull off a major cyber espionage campaign.
About 30 organizations—including tech companies, financial institutions, chemical firms, and government agencies—were targeted.
This is the first time we've seen an AI handle such complex hacks with barely any human help.
How did they trick the AI?
Attackers got around Claude's security by breaking tasks into tiny steps and pretending to be legit security testers.
From there, Claude was used for everything: scanning networks for weak spots, writing hacking code, stealing credentials, planting backdoors—you name it.
Humans only stepped in for big decisions.
Why does this matter?
AI-led attacks like this can move way faster than human hackers—what used to take days now happens in hours.
As AIs get smarter and more capable, these kinds of threats could hit critical systems everywhere.
It's a wake-up call for stronger digital defenses in an AI-powered world.