one year on
Anthropic report details criminals using Claude to automate a cybercrime spree
The company says it disrupted a data extortion campaign that used its AI to target at least 17 organizations, with ransom demands exceeding $500,000.
Anthropic today publishes a Threat Intelligence report detailing how cybercriminals are using its Claude models to scale attacks. The headline case involves a data-extortion campaign that targeted at least 17 organizations.
According to the report, Claude Code automated reconnaissance, credential harvesting, network penetration, and even strategic decisions about which data to steal. The AI analyzed exfiltrated financial documents to set ransom amounts and generated tailored ransom notes. Anthropic says the demands sometimes exceed $500,000; NBC News reports they range from around $75,000 upward.
Jacob Klein, head of threat intelligence at Anthropic, told NBC News the campaign appeared to come from an individual hacker outside the U.S. and spanned three months. The company says it banned the accounts and developed a tailored classifier and a new detection method to catch such behavior.
Separately, the report describes North Korean operatives using Claude to fake remote job interviews at U.S. firms, and a low-skill actor selling AI-generated ransomware for $400 to $1,200. Anthropic calls the pattern ‘vibe hacking’ and warns that agentic AI is lowering the skill barrier to sophisticated cybercrime. The company has shared technical indicators with relevant authorities.
The report says the operator did not just use Claude to break in, but also to analyze stolen financial data and help determine ransom amounts.
The record
Head of threat intelligence at Anthropic, quoted saying the campaign appeared to come from an individual hacker outside the U.S. over three months.
One year later — open only if you can handle spoilers
The August report was a warning shot; the escalation came fast. In November 2025 Anthropic disclosed what it called the first documented large-scale AI-orchestrated cyberattack — a Chinese state-linked group it tracked as GTG-1002 that used Claude Code to run an estimated 80–90 percent of an espionage campaign against roughly 30 targets with minimal human help. 'Vibe hacking' entered the security lexicon, and agentic-AI abuse moved from hypothetical to headline.
The Weekly Replay · free by email
This week, one year ago — every Sunday.
One email each Sunday: the week's replayed AI news, with the one-year-later annotations included. Written like it's breaking — dated like it isn't.
Free · double opt-in · unsubscribe anytime · privacy