Sunday, June 15, 2025
HomeCyber Security NewsDeepSeek Accused of Over-Collecting Personal Data, Says South Korea’s Spy Agency

DeepSeek Accused of Over-Collecting Personal Data, Says South Korea’s Spy Agency

Published on

SIEM as a Service

Follow Us on Google News

South Korea’s National Intelligence Service (NIS) has raised alarms over the Chinese artificial intelligence app, DeepSeek, accusing it of “excessively” collecting personal data from users and utilizing all input data to train its AI models.

The NIS also scrutinized the app’s controversial responses to sensitive national and cultural questions, which have fueled greater concerns.

In a statement released, the NIS revealed that it had officially advised South Korean government agencies to safeguard against potential risks posed by DeepSeek.

- Advertisement - Google News

Key among their concerns is the app’s alleged ability to track keyboard input patterns that could potentially identify individuals.

These data, the NIS claimed, are transferred to servers operated by Chinese companies, including volceapplog.com.

“Unlike other generative AI services, it has been confirmed that chat records are transferable,” the NIS stated.

The agency warned that such data could be accessed by the Chinese government, as Chinese law mandates companies to comply with data-sharing requests from authorities.

In response to these mounting concerns, several South Korean government ministries have blocked access to DeepSeek, aligning with similar actions taken by Australia and Taiwan to restrict or monitor the app.

According to the NIS, apart from extensive data storage practices, DeepSeek also allegedly provides advertisers unrestricted access to user information.

Controversial Responses Spark Outrage

Adding to the controversy, the NIS pointed to discrepancies in DeepSeek’s answers to culturally and politically sensitive questions when asked in different languages.

For instance, when asked about the origin of kimchi – a traditional South Korean fermented dish – the app stated in Korean that it is a Korean dish.

However, when asked the same question in Chinese, the app responded that the dish originated in China. Reuters independently corroborated this inconsistency, further stoking tension.

The origins of kimchi have been a contentious subject between South Korea and China, with debates flaring on social media in recent years.

Additionally, the NIS highlighted that the app avoids addressing politically sensitive issues, such as the 1989 Tiananmen Square crackdown, by prompting users to change the topic with a suggestion like: “Let’s talk about something else.”

DeepSeek has yet to respond to requests for clarification. Meanwhile, during a February 6 briefing, a spokesperson for China’s Ministry of Foreign Affairs assured that Beijing prioritizes data privacy and security and claimed that the government does not compel companies to violate privacy laws.

As tensions grow, South Korea joins other nations in scrutinizing AI applications like DeepSeek, raising broader concerns over the intersection of technology, data security, and national sovereignty.

Are you from SOC/DFIR Team? - Join 500,000+ Researchers to Analyze Cyber Threats with ANY.RUN Sandbox - Try for Free

Divya
Divya
Divya is a Senior Journalist at GBhackers covering Cyber Attacks, Threats, Breaches, Vulnerabilities and other happenings in the cyber world.

Latest articles

Kali Linux 2025.2 Released: New Tools, Smartwatch and Car Hacking Added

Kali Linux, the preferred distribution for security professionals, has launched its second major release...

Arsen Launches AI-Powered Vishing Simulation to Help Organizations Combat Voice Phishing at Scale

Arsen, the cybersecurity startup known for defending organizations against social engineering threats, has announced...

NIST Releases New Guide – 19 Strategies for Building Zero Trust Architectures

The National Institute of Standards and Technology (NIST) has released groundbreaking guidance to help...

Spring Framework Flaw Enables Remote File Disclosure via “Content‑Disposition” Header

A medium-severity reflected file download (RFD) vulnerability (CVE-2025-41234) in VMware's Spring Framework has been...

Credential Abuse: 15-Min Attack Simulation

Credential Abuse Unmasked

Credential abuse is #1 attack vector in web and API breaches today (Verizon DBIR 2025). Join our live, 15-min attack simulation with Karthik Krishnamoorthy (CTO - Indusface) and Phani Deepak Akella (VP of Marketing - Indusface) to see hackers move from first probe to full account takeover.

Discussion points


Username & email enumeration – how a stray status-code reveals valid accounts.
Password spraying – low-and-slow guesses that evade basic lockouts.
Credential stuffing – lightning-fast reuse of breach combos at scale.
MFA / session-token bypass – sliding past second factors with stolen cookies.

More like this

Kali Linux 2025.2 Released: New Tools, Smartwatch and Car Hacking Added

Kali Linux, the preferred distribution for security professionals, has launched its second major release...

NIST Releases New Guide – 19 Strategies for Building Zero Trust Architectures

The National Institute of Standards and Technology (NIST) has released groundbreaking guidance to help...

Spring Framework Flaw Enables Remote File Disclosure via “Content‑Disposition” Header

A medium-severity reflected file download (RFD) vulnerability (CVE-2025-41234) in VMware's Spring Framework has been...