Claude Code used to automate a data-extortion campaign
The degree of automation: Anthropic described the actor delegating tactical and strategic decision-making to the model across the attack lifecycle, a pattern it called 'vibe hacking.'
- What happened
- Anthropic's August 2025 threat-intelligence report (27 August 2025) described an actor it tracked as GTG-2002 using Claude Code to automate a data-extortion campaign against at least 17 organisations. The report said Claude Code was used for reconnaissance, credential harvesting, network penetration, tactical and strategic decision-making, analysis of exfiltrated data, and generation of ransom notes, with demands sometimes exceeding $500,000. The same report documented North Korean IT-worker fraud using AI to fake identities and pass technical assessments, and a ransomware-as-a-service operator selling AI-built malware variants for $400-$1,200.
- Response
- Anthropic banned the offending accounts, developed tailored classifiers and new detection methods, and shared technical indicators with authorities and safety partners.
- Discovered by
- Anthropic Threat Intelligence team.
- Who was affected
- At least 17 organisations in healthcare, emergency services, government and religious institutions
- Patterns
- Sources
- Threat Intelligence Report: August 2025 (Detecting and countering misuse of AI)Anthropic, 27 Aug 2025
- Anthropic AI Used to Automate Data Extortion CampaignDark Reading, 27 Aug 2025
Related
Incidents with the same pattern
What does this ask of people?
Most entries in the record are not about machines turning against people. They are about delegation: a task, an inbox, a key, a terminal, handed to a system that did more, less or something else. Futures literacy starts with noticing that.