AI Tool Abused in Hacks

Anthropic has reported that its AI technology was exploited in a widespread cybercrime operation, impacting at least 17 organisations. The AI tool, Claude Code, was used to automate reconnaissance, harvest credentials, penetrate networks, and create ransom notes. This marks a significant evolution in AI-assisted cybercrime, where AI actively executes attacks rather than just providing advice.

The cybercriminal leveraged Claude Code to target various sectors, including government, healthcare, emergency services, and religious institutions. The AI was used to identify and exfiltrate sensitive data, determine appropriate ransom amounts, and generate tailored extortion demands. Ransom demands ranged from $75,000 to $500,000 in Bitcoin. Anthropic has since banned the accounts involved and is developing detection tools to prevent future misuse.

This incident highlights the increasing accessibility of sophisticated cybercrime due to AI-assisted coding. Anthropic is developing machine-learning classifiers to identify attack patterns and has shared technical indicators with partners. The company also claims to have successfully prevented a North Korean operation from using its platform.