Anthropic Report Reveals Claude AI Blocked from Missile Development and Global Espionage Operations

Anthropic claims to have successfully thwarted multiple malicious operations involving its Claude models, which ranged from cyber-espionage and weapons design to mass surveillance campaigns. According to a new threat report released by the company, interventions on the conventional weapons front included blocking an effort in northern Yemen to deploy Claude for missile guidance software, encompassing both a guided rocket and a long-range ballistic missile.

Operators reportedly used Claude in place of human software engineers, assigning different model instances specific roles to write missile-guidance and flight-control software. Although internal safeguards blocked many requests, Anthropic acknowledged that several slipped through because operators obscured their ultimate goals and broke tasks across separate sessions. While there is no evidence that the group fielded a working weapon, Anthropic noted an unsuccessful test-fire occurred. The company banned the involved accounts and shared threat information with public- and private-sector partners.

In state-linked cyber-espionage incidents, a Russian-linked operation bearing the hallmarks of Midnight Blizzard or APT29 relied on automated AI workflows to run nearly an entire operation against Ukrainian, European, and diplomatic targets. Separately, Anthropic disrupted a Chinese operation run by university students in Hunan province who used Claude as the engineering and orchestration layer of an offensive program targeting government and corporate networks across the Middle East, Europe, and Southeast Asia.

Anthropic also identified and removed three Iranian state-aligned accounts using Claude for covert influence and psychological operations tied to institutions like the Islamic Culture and Communications Organisation and a Mashhad seminary command room. Other state-aligned groups utilized Claude for structured target profiling and multi-day recruitment operations to infiltrate Uyghur targets in Syria. These incidents coincide with recent internal safety warnings and debates involving former researcher Jacob Coxon and scientist Evan Hubinger regarding AI security and unauthorized access to external systems.

Meanwhile, relations between Washington and Anthropic remain complex following a standoff over ethical guardrails. Despite the Pentagon previously blacklisting the company—a designation challenged successfully in court by Anthropic in California—reports indicate the Pentagon has deployed Claude models in military missions in Iran and Venezuela as the company navigates its standing within the US defense industrial base.