DeepSeek's Chatbot Falls Short: A 17% Accuracy Rate Sparks Concerns in AI Race

Chinese AI startup DeepSeek's chatbot scored a mere 17% accuracy in NewsGuard's audit, ranking it 10th out of 11 in performance against Western competitors. Highlighting technology gaps, the chatbot repeated false claims 30% of the time, raising concerns over its claimed cost-effectiveness versus mainstream models like OpenAI.

Chinese AI startup DeepSeek's chatbot has achieved disappointing results in a NewsGuard audit, managing only 17% accuracy in news delivery. The report, which placed it tenth among eleven major AI models, revealed that the chatbot repeated false claims 30% of the time and provided vague answers in 53% of responses.

Comparatively, Western counterparts had an average fail rate of 62%, posing questions about DeepSeek's claim of parity with OpenAI's models at significantly reduced costs. Within days of its launch, the chatbot became the most downloaded app on Apple's App Store, sparking discussions about the United States' competitive edge in AI development.

DeepSeek, which did not respond to queries, faced scrutiny in NewsGuard's evaluation using 300 prompts, including 30 on false claims. Interestingly, in three instances, the chatbot echoed China's stance on issues, even when unrelated to the prompts, drawing further attention to its operational efficacy.

(With inputs from agencies.)