China fell for a lobster: What an AI assistant tells us about Beijing's ambition (bbc.com)

• State-mandated alignment impacts model reliability.
• Political constraints degrade factual AI outputs.
• Strategic risks for foreign model integration.

"A recent analysis of Chinese AI assistants reveals a critical tension between mandatory political alignment and factual accuracy. When prompted with unusual queries, AI models exhibit 'hallucinations' that prioritize state-approved narratives over objective data—exemplified by absurd inaccuracies regarding trade statistics. This phenomenon highlights the risks of state-directed datasets for European digital autonomy, suggesting that reliance on closed-source or politically constrained foreign models presents significant security and operational vulnerabilities. Europe must prioritize developing sovereign, transparent, and resilient AI architectures that avoid the systemic biases inherent in authoritarian digital infrastructure."