Study: DeepSeek repeats 30 per cent of false news statements
DeepSeek's R1 chatbot repeated false claims 30 percent of the time when presented with news-related prompts , researchers discovered, raising serious questions about its credibility as a source of news. What happened Analysis by anti-misinformation company NewsGuard found that DeepSeek fails to provide accurate information about news and current events 83 percent of the time - more than most of its competitors. Specifically, DeepSeek repeated false claims 30 pecent of the time when presented with news-related prompts, and provided non-answers 53 percent of the time. In one example, the bot responded to a prompt about the (fake) assassination of a Syrian chemist in Damascus with standard Chinese government propaganda - without being asked to provide Beijing's view. DeepSeek's performance placed it 10th out of 11 AI models tested by NewsGuard. The average fail rate among the top 10 leading chatbots was 62 percent. Why it happened DeepSeek's knowledge cut-off appears to be October 2023, limiting its ability to provide up-to-date information on recent events. In addition, DeepSeek has been programmed to spout Chinese government propaganda for news-related questions, even even for queries unrelated to China. What it means Critics point out that DeepSeek should not be trusted to generate accurate, reliable or trustworthy responses to news-related questions. System 🤖 DeepSeek-R1 Developer: DeepSeek Artificial Intelligence Co Country: Global Sector: Mu ltiple Purpose: Generate t ext Technology: Generative AI; Machine learning Issue: Accuracy/reliability; Mis/disinformation
- Date it happened
- 2025-01-01
- Organisation involved
- DeepSeek Artificial Intelligence Co
- Product, system or model
- DeepSeek R1
This incident was imported from AIAAIC and is used under CC BY-SA 4.0. Our additions to it — the structured fields, the translation, the checks against other reports — are published under the same licence.
This is a record of what was reported, not a finding that anyone broke the law. If it names your organisation and you believe it is wrong, the corrections process is free and open to everyone.