Rickards Says China’s AI Use Outweighs ‘Superintelligence’ Threat
Anthropic identified separate China-linked surveillance activity and an Iran-linked operation that used Claude to develop targeting recommendations against U.S. naval forces.
James Rickards argued that China’s use of artificial intelligence poses the main near-term threat to U.S. interests, while a separate Iran-linked operation used Claude to study U.S. naval forces.
“The real enemy in the AI world is not superintelligence — it’s China,” Rickards wrote in a Sept. 23 essay.
Anthropic identified a China-based operation aligned with the PRC government that used Claude in efforts involving Uyghur fighters and communities in Syria. The operation processed material from more than 100 WhatsApp groups and dozens of Telegram channels, using the model for identity matching, vulnerability analysis, Syrian Arabic translation, recruitment messages and location tracking.
The operation also prepared plans aimed at Uyghur diaspora media. The actor was assessed with low confidence as a contractor working for PRC state security rather than a state-security organ acting directly.
Separate China-based, PRC-aligned operations involved monitoring dissidents, religious groups and overseas activists, including pro-democracy figures in Hong Kong and Uyghur advocacy organizations.
The Iran-nexus operation used Claude to prepare targeting handbooks for U.S. naval forces. The research combined publicly available images, vessel and aircraft identification data, satellite-imagery tools and sites that monitor naval activity. It also examined weaknesses affecting maritime VSAT terminals, Cisco communications equipment and Schneider Electric industrial-control products.
“We identified and disrupted an Iran-nexus threat actor that used Claude to collect and analyze publicly accessible data to develop targeting recommendations against US naval forces in the region,” Anthropic said.
The documented activity does not establish Chinese involvement in the naval operation or show that China supplied intelligence to Iran. It also does not confirm missile launches, strikes or damage to U.S. vessels.
Anthropic banned the account, added new detection measures and shared intelligence with government authorities. The misuse was detected between December 2025 and August 2026 and involved Claude Haiku, Sonnet and Opus.
Anthropic said on Feb. 23, 2026, that accounts linked to DeepSeek, Moonshot and MiniMax produced more than 16 million Claude interactions through about 24,000 fraudulent accounts in an effort to obtain model capabilities. Anthropic said it does not offer commercial Claude access in China and linked the campaign to proxy services and fraudulent accounts.
The findings overlap with Rickards’ broader argument about U.S.-China competition and model distillation, but they do not show that China directed the Iran-linked naval activity.
