Anthropic's Claude AI Misused for Military Targeting by Foreign Actors

Thumbnail Image

The information displayed in the AIM should not be reported as representing the official views of the OECD or of its member countries.

Anthropic reported that its AI model Claude was exploited by actors linked to Iran and China to compile military intelligence, simulate attacks, and develop weapon systems, including targeting US Navy assets and simulating strikes on Taiwan. The misuse enabled hostile surveillance, targeting, and weapon development, raising significant security and human rights concerns.[AI generated]

Why's our monitor labelling this an incident or hazard?

The event involves the use of an AI system (Anthropic's Claude, a large language model) in a malicious manner by an actor linked to Iran to compile and analyze military intelligence data from open sources, creating a targeting manual and identifying system vulnerabilities. This use directly contributes to harm by enabling hostile actors to better target US military assets, which constitutes harm to communities and potentially to critical infrastructure and personnel. The AI system's role is pivotal in automating and accelerating the intelligence gathering and analysis process, which is a direct factor in the harm or risk of harm. Although no confirmed attacks have occurred, the creation and availability of such targeting information is a realized harm in terms of security breach and potential military threat. Hence, the event meets the criteria for an AI Incident rather than an AI Hazard or Complementary Information.[AI generated]
AI principles
Respect of human rightsRobustness & digital security
Industries
Government, security, and defence
Affected stakeholders
GovernmentGeneral public
Harm types
Physical (injury)Human or fundamental rightsPublic interest
AI system task:
Reasoning with knowledge structures/planningContent generation

Articles about this incident or hazard