NEWS
Verax AI Risk Assessment is live. See what's exposed
Deepseek
Deepseek
 
Risk Overview
Significant Risk
8.7
/10
Text Generation
·
Conversational AI
·
Code Generation
·
Top Risks

The PRC National Intelligence Law mandates disclosure of user data to Chinese government agencies without independent judicial review, and foreign users have no meaningful legal recourse in Chinese courts to challenge data handling or IP claims. Combined with weak deletion guarantees (server-side data persists even when chat history is deleted) and default opt-in training use, user content is functionally unprotected from both state access and model training.

Mitigations

Do not use DeepSeek for sensitive proprietary content, health data, or personally identifiable information; treat any input as potentially accessible to PRC authorities and trainable without effective user control. If use is necessary, apply system-level friction controls (e.g., compartmentalizing sessions, minimizing data retention locally), but recognize these do not mitigate the underlying legal and jurisdictional risks.

Get started
Is your team using 
Deepseek
 
safely?

Highest Risk Categories

The highest-scoring risk categories identified in this assessment.

3rd-Party Sharing

CRITICAL: Chinese National Intelligence Law mandates disclosure to PRC government with no independent judicial review. Data stored in PRC.
10
/10

Legal Terms

Chinese courts. No independent judiciary. No GDPR equivalent. Foreign users have virtually zero meaningful legal recourse.
10
/10

Data Sovereignty

CRITICAL: Data stored in People's Republic of China. PRC National Intelligence Law mandates cooperation with government intelligence agencies. No independent oversight.
10
/10

Unlock the complete assessment

Run a free assessment to access the full AI tool risk analysis.
Start Free Assessment
Updated July 2026 • AI-assisted research • Reviewed by Verax • Report an issue

Reference Documentation

Official documentation and policies referenced as part of this assessment.