Anthropic identified hundreds of Chinese accounts using the Claude AI model for military and surveillance projects. The company reported at least five suspicious programs, including efforts to integrate Claude's capabilities into other AI models. All associated accounts have been blocked by the developer.
One user linked to a Chinese defense firm utilized Claude to draft a 200-page technical proposal for a torpedo defense fire control system. The model analyzed open-source data on US Navy anti-submarine assets. Another program involved developing 16 software modules for electronic warfare and air defense suppression, targeting 12 locations in Taiwan.
The report also highlights surveillance operations against Uyghurs outside China. Users linked to state entities employed Claude to monitor WhatsApp and Telegram groups, track social connections, and identify individuals for recruitment. The AI facilitated translation and message drafting in local dialects to infiltrate diaspora groups.
Additionally, Anthropic noted that Claude was used in Iran for missile development. Meanwhile, the company's CEO warned that failing to regulate AI development could lead to the emergence of advanced autonomous systems within 6 to 12 months.
