Anthropic Disrupts AI-Powered Cyberattacks Automating Theft and Extortion Across Critical Sectors

Anthropic on Wednesday revealed that it disrupted a sophisticated operation that weaponized its AI-powered chatbot Claude to conduct large-scale theft and extortion of personal data in July 2025. The actor targeted at least 17 distinct organisations, including in healthcare, the emergency services, government, and religious institutions, with threats to expose stolen data rather than encrypt it, and ransom demands that sometimes exceeded $500,000.

The attack used Claude Code on Kali Linux as a comprehensive platform, embedding operational instructions in a CLAUDE[.]md file to provide persistent context for every interaction, and employed AI to automate reconnaissance, credential harvesting, and network penetration.

The activity, codenamed GTG-2002, reportedly used Claude to decide which data to exfiltrate and to craft bespoke ransom notes and multi-tier extortion strategies based on data analysis, with ransom amounts ranging from $75,000 to $500,000 in Bitcoin. According to Anthropic, the misuse extended to profiling victims, monetising stolen data, and evading detection with customized malware tooling, prompting the company to develop a classifier to screen for similar behaviour and share indicators with key partners.