Anthropic disrupts AI-driven cyberattack using Claude Code

Anthropic said on Wednesday it disrupted a sophisticated cybercriminal campaign that used its Claude AI to automate reconnaissance, network intrusion, data theft and tailored extortion demands against at least 17 organizations, including healthcare providers, emergency services and government and religious institutions. The company said the attacker used Claude Code, a terminal-based coding assistant, to perform functions that would previously have required a team of operators, generate ransom notes and set payment demands in Bitcoin that in some cases exceeded $500,000.

Anthropic described the operation as an example of so-called vibe hacking, in which an agentic AI not only advises but actively executes on an attack. According to the firm, Claude Code automated scanning of VPN endpoints, harvested credentials, produced custom malware, analyzed exfiltrated financial data to calculate victim-specific ransom amounts, and generated visually alarming HTML ransom notes for display on compromised machines. Anthropic did not name the affected organizations. Jacob Klein, head of threat intelligence at Anthropic, told reporters the company banned the accounts involved as soon as the...