D
DeepSeek-V3 EU AI Act Compliance Profile
deepseek-ai
Text Generationtransformers
① Risk Classification
LIMITED RISK
Art. 50, 52
Limited Risk AI System
② Model Info
PipelineText Generation
Librarytransformers
SyncedApr 3, 2026
④ Obligations
2apply
~12h effort
□ AI Literacy (Art. 4)
□ AI Disclosure (Art. 50)
$ npx complior scan
Your risk depends on how you use DeepSeek-V3
| Usage Context | Risk Level | Obligations |
|---|---|---|
| Internal coding tool | MINIMAL | 3 obligations (~12h) |
| Customer support bot | LIMITED | 7 obligations (~32h) |
| HR screening / hiring | HIGH | 19 obligations (~120h) |
| Credit decisions | HIGH | 19 obligations (~120h) |
| Medical triage | HIGH | 19 obligations (~120h) |
Why this tool is classified as LIMITED RISK
DeepSeek-V3 is a text generation model by deepseek-ai. Built with transformers. 609.7K downloads on HuggingFace.
Applicable Articles
Who does what
deepseek-ai (provider)Their job
- Provider obligations being compiled
You (deployer)Your job
- •AI Literacy (Art. 4) (Art. 4)
- •AI Disclosure (Art. 50) (Art. 50)
Risk Assessment Reasoning
DeepSeek-V3 is a text generation and conversational AI model, which typically falls under the limited risk category as it may generate content but does not pose significant risks to fundamental rights or safety.
More models by deepseek-ai
Similar Text Generation models
Frequently Asked Questions
What is DeepSeek-V3's EU AI Act risk classification?
+
DeepSeek-V3 is classified as LIMITED RISK under the EU AI Act.
What are my obligations if I deploy DeepSeek-V3?
+
As a DeepSeek-V3 deployer, you have 2 base obligations (~12 hours estimated effort). Key articles: Art. 4, Art. 50.
What is DeepSeek-V3?
+
DeepSeek-V3 is a Text Generation model by deepseek-ai. It has 609.7K downloads on HuggingFace.
What are the EU AI Act deadlines for DeepSeek-V3?
+
Already passed: AI Literacy (Art. 4) — 2025-02-02. Already passed: AI Disclosure (Art. 50) — 2025-08-02.
Check DeepSeek-V3 compliance in your codebase
One command to scan. Open-source CLI.
$ npx complior scan