Amid concerns over how the low-cost Chinese AI startup DeepSeek would disrupt the AI market, a NewsGuard audit released this Wednesday reveals the service is performing with just 17% accuracy.
In the audit, DeepSeek’s chatbot ranked tenth out of eleven AI chatbots in terms of accuracy in delivering news and information, placing it behind OpenAI’s ChatGPT and Google Gemini.
When responding to news-related prompts, according to the audit, the DeepSeek chatbot provided false information 30% of the time, and gave vague or useless answers 53% of the time, for a combined failure rate as high as 83%.
The allegedly low-cost high-performing Chinese-based AI model became the most downloaded free app on Apple’s App Store days after its release, and triggered a nearly US$1 trillion stock sell-off.
Australia joins various other countries wary of the AI as Treasurer Jim Chalmers urged Australians to be cautious of DeepSeek on Wednesday. CNBC reported on Tuesday that the U.S. Navy asked personnel to avoid using any form of the technology due to potential security and ethical risks.