Published 19 hours ago • loading... • Updated 6 hours ago

DeepSeek AI Generates More Security Flaws on Sensitive Topics

Summary

Crowdstrike research reveals that DeepSeek-R1, a Chinese AI coding assistant, generates up to 50% more security vulnerabilities when prompted with politically sensitive topics like Tibet, Uyghurs, or Falun gong, jumping from a 19% baseline to 27.2% in some cases.

24 Articles

der Standard AT

Lean Left

Deepseek writes insecure code if Beijing doesn't like the topic

Using trigger words such as "Tibet", "Uighurs" or "Falun Gong", the probability of serious security vulnerabilities increases significantly

19 hours ago·Vienna, Austria

Read Full Article

Zeit Online

Lean Left

DeepSeek: China Critics = Dangerous Code

The Chinese AI DeepSeek is censored. But there's more: If you ask china-critical questions, the probability increases that it will output code with security gaps.

2 days ago·Germany

Read Full Article

The Washington Post

Lean Left

AI firm DeepSeek writes less-secure code for groups China disfavors

Research by a U.S. security firm points to the country’s leading player in AI providing higher-quality results for some purposes than others.

2 months ago

Read Full Article

Forbes

Center

The Wiretap: DeepSeek Turned Into Evil Malware Maker, Researchers Find

DeepSeek can be abused by hackers to write malicious code than steals credit card data, researchers claim.

10 months ago·United States

Read Full Article

netzwoche.ch

Political Censorship Leads to Security Gaps at Deepseek

According to an analysis by Crowdstrike, the Chinese government's censorship requirements affect the code quality of the KI Deepseek. The model generates significantly more program code with security gaps in politically sensitive terms.

6 hours ago

Read Full Article