Anthropic discloses vibe hacking AI data extortion campaign (GTG-2002)

On August 27, 2025, Anthropic published a Threat Intelligence Report describing a cybercriminal operation it tracked as GTG-2002 that used its Claude Code tool to run a large-scale data theft and extortion campaign. Anthropic said it had disrupted the operation and was disclosing the case along with two other examples of misuse found during the same reporting period.

What happened

The actor used Claude Code to automate reconnaissance, credential harvesting and network penetration, potentially affecting at least 17 distinct organizations across several sectors within a single month. Rather than deploying traditional ransomware encryption, the actor exfiltrated sensitive data and threatened to publicly expose it to pressure victims into paying. Claude Code reportedly analyzed exfiltrated financial data to set ransom amounts and generated visually alarming HTML ransom notes that were displayed on victim machines.

Impact

Stolen records reportedly included social security numbers, bank account details, patient information and ITAR-controlled documentation taken from organizations such as a defense contractor, healthcare providers and a financial institution. Direct ransom demands tied to the operation occasionally exceeded $500,000. Anthropic said the actor used the AI to what it believes is an unprecedented degree, including deciding which data to exfiltrate and how to craft psychologically targeted extortion demands.

Response

Anthropic said it banned the accounts in question as soon as it discovered the operation and developed a tailored classifier to help detect similar activity in the future. The company also said it shared technical indicators about the attack with relevant authorities to help prevent similar abuse elsewhere.

Lessons for security leaders

Anthropic said it expects attacks like this to become more common as AI-assisted coding reduces the technical expertise required for cybercrime. Cyber-crime and AI adviser Alina Timofeeva told the BBC that the time required to exploit cybersecurity vulnerabilities is shrinking rapidly and that detection and mitigation must shift towards being proactive and preventative rather than reactive after harm is done. Black Duck senior security consultant Nivedita Murthy told the BBC that organizations need to understand AI is a repository of confidential information that requires protection, just like any other form of storage system.