FAQs
What is DeepSeek AI?
Frequently Asked Questions #DeepSeek
This FAQ unpacks what makes DeepSeek unique, why it’s raising global concerns, and how Tenable’s AI Aware capability empowers organizations to detect and manage unauthorized AI usage. Learn how DeepSeek compares to frontier models, explore potential data privacy and security issues, and understand Tenable’s approach to safeguarding against emerging AI/LLM threats.
What is DeepSeek?
DeepSeek refers to a large language model (LLM) developed by the Chinese company DeepSeek, founded in 2023 by Liang Wenfeng.
What is a Large Language Model (LLM)?
A large language model (LLM) is a machine-learning model trained on vast amounts of text data, enabling it to generate human-like responses to user queries.
Why is there so much interest in the DeepSeek LLM?
In January 2025, DeepSeek released two open-source models, DeepSeek-v3 and DeepSeek-R1. These models generated interest because:
- They are open-source, allowing anyone to download and run them locally.
- They were reportedly trained using less-powerful hardware, representing a potential breakthrough in AI efficiency.
What makes DeepSeek different from other LLMs?
- Benchmark tests show DeepSeek-R1 performs similarly to frontier models like OpenAI’s o1.
- Unlike many competitors, DeepSeek has open-sourced its models, making them more accessible.
- DeepSeek’s Chain of Thought (CoT) reasoning approach provides more detailed insights into its decision-making process.
DeepSeek was created by a Chinese company. Is it safe to use?
Running the open-source version locally reduces certain risks, but utilizing DeepSeek’s website or mobile applications carries potential privacy and security concerns.
What are the concerns surrounding using DeepSeek?
- Data Privacy: User data is stored on servers in China, outside the protection of GDPR and other global privacy regulations. DeepSeek's privacy policy.
- Third-Party Data Sharing: DeepSeek’s web analytics are processed via Baidu, and network/device data is shared with ByteDance.
- Security Risks: Weak model guardrails make DeepSeek susceptible to jailbreaks. Cisco, Palo Alto Networks, and Wiz have reported security flaws.
- Censorship & Bias: Enkrypt AI identified high censorship rates and pro-China biases in DeepSeek’s responses.
Has DeepSeek been banned anywhere or is it being reviewed for a potential ban?
As of February 13, 2025, several governments and agencies have banned or are reviewing DeepSeek, including:
- Countries: Italy, Taiwan, Australia, South Korea, India, and United States (various federal agencies).
- U.S. State: Texas, New York, and Virginia.
- Agencies: NASA, Pentagon, U.S. Navy, U.S. Congress, Defense Information Systems Agency (DISA).
To learn more, visit:
- Details on Federal Agency DeepSeek response
- Tenable's Threat Research on DeepSeek blog
- Tenable's Threat Research using DeepSeek to create keylogger and ransomware blog
How does Tenable help secure organizations against unauthorized use of DeepSeek? #Tenable AI Aware
For Tenable Vulnerability Management customers, the following AI Aware plugins have been added to detect DeepSeek and enforce security policies:
These detections assist organizations in documenting and controlling unauthorized DeepSeek usage, identifying and mitigating DeepSeek vulnerabilities, and aligning DeepSeek app development with security and governance policies.
What is AI Aware?
Tenable AI Aware is a capability within Tenable Vulnerability Management and Tenable Security Center designed to help monitor and manage AI-related risks. It uses advanced detection technologies to surface unauthorized AI solutions, identify AI vulnerabilities, and detect unexpected AI or Large Language Model (LLM) development within your organization.
To learn more, visit
For further inquiries, please get in touch with Damien Lim, [email protected].
- AI
- Cloud
- Vulnerability Management
- Tenable Cloud Security
- Tenable Vulnerability Management
