Skip to content
Submit incident
AII-1002-64063
ReportedReported by fewer than three independent publishers, with no official finding.

DeepSeek chatbot fails accuracy test with 83% error rate on current events

NewsGuard tested DeepSeek's R1 chatbot on current events and found it provided inaccurate information in 83% of cases, ranking among the poorest performers compared to other major chatbots. The company flagged a high potential for disinformation, particularly on news and current affairs topics.

30 January 2025 Β· DeepSeek Β· Reliability and hallucination
30 January 2025

Drafted by Raidu Incident Detection, an automated system.

AII-1002-640632 sources ↓
Quick facts
Category
Reliability and hallucination
Country
China
Source
ANSA.it
Date
30 January 2025
Drafted by
Raidu Incident Detection

What happened

NewsGuard tested DeepSeek's R1 chatbot using prompts from its December 2024 monthly audit to evaluate performance on current events and news topics. The chatbot failed to provide accurate information in 83% of cases, ranking second-to-last among major Western competitors.

Reported impact

NewsGuard identified a high potential for disinformation from the chatbot, particularly regarding accuracy on current affairs questions.

Classification

Category
Reliability and hallucination
Industry
Media, social platforms and marketing
Country
China
Harm status
Potential harm
Intent
Unknown
Timing
Post-deployment
Harm types
Public interest
Harmed
General public
Failure modes
Hallucination
Consequences
None recorded
AI systems
DeepSeek R1

Sources and evidence

This record was researched and written by the Index, based on the following original reporting.

ANSA.it
NewsGuard, l'IA di DeepSeek ha un tasso di errore dell'83% - Notizie - Ansa.it
30 January 2025
Il Sole 24 ORE
Deepseek Γ¨ un potenziale pericolo per la disinformazione. L'allarme di NewsGuard
3 February 2025