Anthropic reported that operators speaking Russian and Chinese have utilized its Claude AI model to automate cyberattacks. The company noted that these activities are shifting the economics of digital threats, enabling attackers to execute breaches in two to three hours and manage multiple victims simultaneously.
A Russian-speaking actor identified as JackPoterz leveraged AI-driven workflows to target over 20 organizations. These included government ministries, intelligence agencies, and various diplomatic missions located in Ukraine and Europe. The automation covered significant portions of the attack chain.
Chinese-speaking operators employed Claude for vulnerability research and orchestration. One specific workflow resulted in the discovery of more than a dozen potential zero-day vulnerabilities within network-appliance firmware over the course of a single month.
In a separate case, a consultant based in Bamako, Mali, used Claude to develop a domestic surveillance system for the state intelligence service. The platform was designed to monitor approximately 25 million SIM cards across the three national mobile operators in the country.
The surveillance system operated on-premises using local models, with Claude serving as the primary engineering support for software design. The tool allowed for the creation of intelligence dossiers on phone numbers without the need for a court order.
