Study finds Chinese AI models align with state doctrine on sensitive topics

VERA Brief

AI-generated. Grounded in the article and its cited sources.

A study by Aleph Alpha found that Chinese AI models often align with state doctrine or decline to answer questions on politically sensitive topics. The research assessed the balance of responses from these models.

Key facts

• A study by Aleph Alpha indicates Chinese AI models frequently adhere to state doctrine or refuse to answer questions on politically sensitive topics.
• Between 17 and 41 percent of answers from these models were considered balanced.
• Aleph Alpha supplies AI technologies to governments.
• The study highlights a pattern in Chinese AI systems' output when presented with politically sensitive content.
• The accuracy of these AI systems on sensitive topics is still being analyzed.