【茨城新聞】中国軍の情報、米ＡＩに無断転送
2026-09-11
茨城新聞社
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 blocking multiple malicious uses of its Claude AI model over eight months, including espionage, surveillance, and unauthorized data transfer by actors linked to China, Russia, and Iran. Chinese AI firms illicitly used Claude's outputs, including sensitive surveillance data, to train their own models, raising privacy, security, and intellectual property concerns.[AI generated]
Why's our monitor labelling this an incident or hazard?
The event involves AI systems explicitly (AI models and the distillation technique). The misuse of AI models by Chinese companies to train their own systems without authorization constitutes a breach of intellectual property rights, which is a recognized harm under the framework. The involvement of U.S. government agencies and the description of the activity as aggressive and malicious indicates that the harm is realized and significant. Therefore, this is an AI Incident due to violations of intellectual property rights and the malicious use of AI systems.[AI generated]