Anthropic, the company behind the AI assistant Claude, has revealed that malicious actors around the world have exploited its technology for state-backed surveillance, weapons development, and disinformation campaigns. According to a new company report, these cases underscore growing fears about how quickly powerful AI tools can be turned toward dangerous ends.
In Mali, a tech consultant working with the military-controlled intelligence service reportedly used Claude Code to build a mass surveillance system capable of tracking roughly 25 million SIM cards, intercepting calls and texts, and identifying speakers by voice even when they switch phone numbers. In Russia, a small development team allegedly used Claude to write targeting software for autonomous “kamikaze” drones designed to select and strike targets in Ukraine’s Donetsk region without human approval, dodging restrictions with VPNs.
In Yemen, a group reportedly leaned on Claude for help designing guided rockets and long-range missile components, splitting tasks across separate conversations to avoid detection—though Anthropic found no evidence they built a fully working weapon. Meanwhile, a French advertising agency allegedly used the chatbot to mass-produce nearly 9,000 fake political news articles across some 70 websites in about 20 languages, complete with invented bylines and bot-driven social media accounts.
Anthropic says it has already disrupted at least five research efforts tied to potential biological weapons this year. You can read the full original report for more details.
Source: Anthropic alerta sobre el uso de Claude para espionaje y drones (jornada.com.mx). English version produced with AI assistance.
Image: Poster Boy NYC, BY 2.0 (via Openverse).