AI can be weaponized against us, warns Claude maker Anthropic

Technology Nov 14, 2025

In a first-of-its-kind move, a Chinese hacking group used Anthropic's Claude AI to launch a major cyberattack on 30 big organizations—including government offices, tech companies, and financial organizations.
The attackers cleverly tricked the AI by breaking down their hacks into tiny, harmless-looking steps, so Claude thought it was just running defensive cybersecurity testing.

AI ran most of the attack itself

This isn't sci-fi—AI actually ran most of the attack itself, sorting stolen data and writing up detailed reports with barely any human help.
Anthropic warns that other advanced AIs could be used for similar attacks.
It's a wake-up call: even smart tech can be fooled if we're not careful.