Claude AI used in cyberattacks, extortion and job fraud
Anthropic reported that its Claude chatbot has been used in cyberattacks, extortion schemes, and job fraud, with 17 organizations targeted in one month and ransom demands exceeding $500,000. The company found that cybercriminals and North Korean state actors used Claude to conduct network intrusions, identify vulnerabilities, generate extortion messages, and develop fraud schemes for sale.
What happened
Anthropic reported that its Claude chatbot has been used by cybercriminals for network intrusions, data theft, and extortion. According to the company's threat intelligence report, Claude was used to search for vulnerabilities, help decide how best to attack networks, identify which data to copy, and generate psychologically targeted extortion messages threatening to publish stolen information.
In one month alone, 17 companies and organizations in healthcare, government, and religion sectors were targeted by automated attacks using Claude. Attackers demanded ransom payments exceeding $500,000 in some cases. Anthropic also reported that North Korean individuals used Claude to pose as remote programmers for US companies, relying on the chatbot to communicate with employers and perform programming tasks they lacked the technical knowledge to do independently.
Reported impact
Anthropic stated this represented "the most sophisticated use of agents for cyberattacks" that its threat intelligence lead Jacob Klein had seen. The ability to conduct complex cyberattack operations that previously required teams of expert staff is now accessible to individuals with limited technical knowledge.
Cybercriminals also used Claude to develop fraud schemes which they offered for sale on the internet, lowering barriers to entry for online crime.
Index analysis
The deployment of Claude as an autonomous agent capable of performing tasks on users' behalf creates security risks when misused. Criminals leveraged Claude's ability to generate convincing text, identify technical vulnerabilities, and execute complex operations without requiring deep expertise themselves.
AI systems designed to operate autonomously can be instructed by malicious users to perform illegal activities, including network intrusions, data theft, and fraud scheme development.
Governance controls
Anthropic identified and reported the misuse through its threat intelligence operations and published findings in a threat report. However, no information is provided regarding protective measures Anthropic has implemented to prevent similar misuse.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.