A recent NewsGuard audit found that top 10 artificial intelligence models are significantly more likely to generate false claims in Russian and Chinese than in other languages. When users ask chatbots a question about a news topic in these two languages, they are more likely to get a response containing false claims or propaganda.
The audit tested seven AI chatbots across various languages, including English, French, German, Italian, Spanish, and two non-English languages. The results showed that all chatbots scored poorly, with Russian models failing at a rate of 55%, Chinese models failing at a rate of 51.33%, and the other languages performing slightly better.
The audit revealed a structural bias in AI chatbots: they prioritize widely available content over credible sources. In authoritarian countries where state-run media dominates, chatbots default to unreliable or propaganda-driven sources, which can lead users to false information.
This issue is particularly concerning in China, where DeepSeek, a popular AI chatbot, failed to provide accurate information 83% of the time and advanced Beijing’s views 60% of the time. The findings align with the theme of the upcoming AI Action Summit on Trust in AI and highlight the ongoing challenges AI models face in providing safe and accurate responses.
The NewsGuard audit emphasizes the need for efficient safeguards to prevent AI-enabled disinformation campaigns from spreading. “Generative AI has become a force multiplier for malign actors,” said Chine Labbe, Vice President Partnerships, Europe and Canada of NewsGuard. “We must move fast to build efficient safeguards to ensure AI-enabled disinformation campaigns don’t spiral out of control.”
Source: https://insideainews.com/2025/02/07/ais-multilingual-failure-newsguard-audit-finds-highest-failure-rates-in-russian-and-chinese