Storia in 2 fonti

'Multi-part case study on China’s media' finds that AI models can't hallucinate away Chinese censorship | Fortune

New research suggests Chinese state media and authoritarian speech restrictions can shape the answers produced by leading U.S. AI models

Raccontata da

hindustantimes.com

fortune.com

Confronto fonti

2 prospettive sulla stessa storia

AI · summaries

fortune.comStai leggendo7 g fa

'Multi-part case study on China’s media' finds that AI models can't hallucinate away Chinese censorship |…

Nature: 3M state-media docs in AI training data; Claude, GPT-4o reproduce Beijing phrases in 3–10% on China. Meta audit shows US models refuse political criticism 34% more often in authoritarian countries, signaling authoritarian information controls reach globally-deployed AI systems.

originale

hindustantimes.com7 g fa

Chinese censorship is leaking into answers from American AI

Claude, GPT, and Gemini refuse Xi criticism but create US/UK protest content; state-controlled media in training data drives the asymmetric bias. Tech leaders: LLM bias toward autocracies is systematic, not tunable; model selection requires source transparency before deployment in sensitive contexts.

'Multi-part case study on China’s media' finds that AI models can't hallucinate away Chinese censorship |…

Chinese censorship is leaking into answers from American AI

Timeline cronologica

Chinese censorship is leaking into answers from American AI

'Multi-part case study on China’s media' finds that AI models can't hallucinate away Chinese censorship | Fortune