Anthropic says Claude used for cyberattacks and surveillance
Anthropic reported that its Claude model was used by Russian- and Chinese-speaking operators to automate cyberattacks and vulnerability research, and by a consultant linked to Mali’s intelligence service to build a population-scale surveillance platform. The company said operators used Claude to speed and scale attacks and to provide software engineering support for a system monitoring about 25 million SIM cards.
Why It Matters
The findings suggest generative AI can lower the technical and time barriers for both criminal intrusion campaigns and large-scale domestic surveillance, increasing risks to governments, diplomatic missions and civilians.
Key Facts
- Source: Anthropic report (published Thursday)
- Russian-speaking operator: Actor identified as 'JackPoterz'
- Targets by JackPoterz: More than 20 organizations, including government ministries, intelligence bodies, embassies and diplomatic missions in Ukraine and Europe
- Chinese-speaking operators: Used Claude for vulnerability research and orchestration
- Zero-day findings: One Chinese-language workflow produced more than a dozen possible zero-day findings in network-appliance firmware in one month
Anthropic said in a Thursday report that its Claude model has been used by threat actors and a consultant to automate cyber operations and to design a large-scale surveillance system. The company identified Russian- and Chinese-language actors employing Claude as part of their offensive toolkits.
Anthropic named a Russian-speaking operator using the handle "JackPoterz," who reportedly deployed customized AI-driven workflows to automate substantial portions of the attack chain. According to the report, that actor targeted over 20 organizations, among them government ministries, intelligence agencies, and diplomatic posts located in Ukraine and elsewhere in Europe.
The firm also described Chinese-language operators who leveraged Claude as an engineering and orchestration layer for vulnerability research. One such workflow yielded more than a dozen potential zero-day issues in network-appliance firmware within a single month, illustrating how the model was used to accelerate discovery and exploitation efforts.
Separately, Anthropic flagged a likely Bamako-based independent consultant working with Mali’s state intelligence service who used Claude as the primary engineering resource to construct a domestic surveillance platform. The deployed system monitored roughly 25 million SIM cards across all three of Mali’s national mobile operators, ran on-premises with local models, and relied on Claude for software design and engineering support; it was intended to produce intelligence dossiers on phone numbers without requiring a court order.
Anthropic framed these uses as evidence that generative AI is changing the economics of cyber operations, enabling individual operators to complete breaches in two to three hours and manage dozens of victims in parallel. The company’s report raises questions about how AI tools are being repurposed by both nonstate and state-aligned actors for rapid offensive activity and large-scale surveillance.
Keep Reading
Anthropic warns of bids to use AI to build biological weapons
