DeepSeek
by DeepSeek
LLM provider. Italy/EU privacy probes, US security review, training-data concerns (2025).
Risk Score: 5/100 (Low) · 2+ incidents · Legal 15 · Safety 5 · Privacy 0 · Regulatory 0 · Security 0
Risk Score
Apr 27, 2026
Risk Score Breakdown
Legal Risk
Court cases & lawsuits
15/100
Safety Risk
Incidents & harm events
5/100
Privacy Risk
Breaches & GDPR actions
0/100
Regulatory Risk
FTC, EU enforcement
0/100
Security Risk
CVEs & vulnerabilities
0/100
Incident Timeline
2 total incidents · showing 2 most recent
Apr 2026
Anthropic said DeepSeek and other Chinese AI developers illicitly extracted Anthropic model outputs to improve competing AI products. The allegation raises concerns about intellectual property protection for AI developers.
Oct 2025
A civil rights lawsuit alleging disability-rights violations was filed in the U.S. District Court for the Eastern District of New York (1:25-cv-05691) involving Anthropic, Google LLC, Samsung Electronics, and OpenAI. The parties named in the case are directly affected.
Frequently Asked Questions
What is DeepSeek's AI risk score?
DeepSeek has an AI Risk Score of 5/100 (Low Risk). This score is calculated from 2+ documented public incidents across legal, safety, privacy, regulatory, and security categories.
Is DeepSeek safe to use?
DeepSeek by DeepSeek has a low risk profile based on public data. Organizations should review the full incident list and conduct their own due diligence. This score does not constitute legal advice.
Does DeepSeek have lawsuits?
Yes — our public records show 1 court case(s) for DeepSeek, including: Court Case: Chernets v. Google LLC.
How is the AI Risk Score calculated?
Scores are weighted across 5 categories: Legal (25%), Safety (25%), Privacy (20%), Regulatory (15%), Security (15%). Each incident is scored by severity and type, then decayed based on age. Active lawsuits and fatal incidents do not decay.
Stay ahead of AI risk
Get alerts when DeepSeek risk score changes
New lawsuits, breaches, and regulatory actions — delivered to your inbox.