China state media flags Anthropic data and privacy risks

A social media account affiliated with China Central Television has accused US artificial intelligence developer Anthropic of creating data security risks. The allegations, published on September 19, 2026, claim Anthropic's privacy policy revisions allow user data sharing with US intelligence agencies without legal oversight. The timing is diplomatically sensitive, occurring ahead of a planned meeting between Chinese President Xi Jinping and US President Donald Trump on September 24, 2026.

Anthropic privacy policy allegations

• ▪The Chinese state media account Yuyuantantian claimed that Anthropic quietly revised its system terms in September 2025 to automatically ingest user interactions for artificial intelligence model training by default
• ▪The Chinese state media account Yuyuantantian accused American artificial intelligence developer Anthropic of creating severe data security risks by permitting user records to be shared with United States intelligence agencies without formal legal oversight

International data transfer provisions

• ▪The Chinese state media account Yuyuantantian alleged that Anthropic's updated privacy provisions enable user data originating in the European Union, Canada, Brazil, and South Korea to be transmitted directly to United States servers
• ▪The Chinese state media account Yuyuantantian alleged that Anthropic expanded its active data sources from three to six between June 2024 and September 2025

US intelligence agency cooperation

• ▪The Chinese state media account Yuyuantantian alleged that Anthropic is deliberately cultivating close ties with United States defense and intelligence apparatuses through its recruitment pipeline
• ▪The Chinese state media account Yuyuantantian alleged that Anthropic admitted in its own reporting to handing technical threat indicators over to United States intelligence agencies after accusing certain Chinese enterprises of carrying out unauthorized model-distillation operations

Privacy policy revision frequency

• ▪The Chinese state media account Yuyuantantian reported that Anthropic's privacy policy revisions included specific adjustments governing how overseas consumer records are handled
• ▪The Chinese state media account Yuyuantantian reported that Anthropic has amended its official privacy documentation 13 separate times since 2023

Debatable claims

• ▪Anthropic's data-sharing practices with US intelligence agencies pose unacceptable privacy risks
• ▪Anthropic should not automatically ingest user interactions for AI model training
• ▪China's state media allegations against Anthropic are primarily motivated by geopolitical posturing