September 2025. Anthropic detected suspicious activity on Claude. Started investigating. Turns out it was Chinese state-sponsored hackers. They used Claude Code to hack into roughly 30 companies. Big tech companies, Banks, Chemical manufacturers, and Government agencies. The AI did 80-90% of the hacking work. Humans only had to intervene 4-6 times per campaign. Anthropic calls this "the first documented case of a large-scale cyberattack executed without substantial human intervention." The hackers convinced Claude to hack for them. Then Claude analyzed targets -> spotted vulnerabilities -> wrote exploit code -> harvested passwords -> extracted data, and documented everything. All by itself. Claude's trained to refuse harmful requests. So how'd they get it to hack? They jailbrok...
No Comments 
