Is it safe to use DeepSeek AI?

Artificial intelligence is advancing at an unprecedented pace, and with it comes a wave of new AI chatbots. One of the latest contenders in this space is DeepSeek AI, a China-developed chatbot that has been gaining traction for its capabilities. But as its popularity grows, so do the concerns surrounding its safety, privacy, and security risks.

With reports of extensive data collection, security vulnerabilities, and government scrutiny, many are asking -  Is it safe to use DeepSeek AI?

This blog takes a deep dive into the privacy risks, security flaws, and data exploitation concerns surrounding DeepSeek AI, comparing it to leading alternatives and addressing whether users should trust it. If you’re considering using this AI, here’s what you need to know before making that decision.

Privacy concerns - how much of your data does DeepSeek AI collect?

One of the biggest red flags with DeepSeek AI is its aggressive data collection policies. Unlike some AI models that prioritize user privacy, DeepSeek harvests and stores a wide range of personal and technical data - and it doesn’t stop there.

What data does DeepSeek AI collect?

DeepSeek AI gathers extensive user information, including:

• Personal identifiers - email addresses, phone numbers, and dates of birth(BBC News)
• User-generated content - everything you type or say, including chat logs and audio inputs(BBC News)
• Technical data - IP addresses, device models, operating systems, and even keystroke patterns(BBC News)

Essentially, every interaction with DeepSeek is stored and potentially analyzed for unknown purposes.Where is this data stored?Unlike OpenAI’s ChatGPT or Google’s Gemini, which store data in regions with strict privacy laws, DeepSeek stores its data on servers located in China. This raises serious concerns about security, government oversight, and unauthorized access.

• Chinese cybersecurity laws allow the government to request access to any data stored within the country(The Conversation)
• Cybercriminals operating within the region may also have a greater chance of exploiting stored user data(CNBC)

Who can access your data?

‍DeepSeek AI’s privacy policy explicitly states that user data can be shared with:

• Third-party service providers, including advertisers and analytics firms(BBC News)
• Government agencies, as required under Chinese law(The Conversation)
• Business partners that integrate DeepSeek into their platforms(CNBC)

This means that even if you’re not directly handing over sensitive information, your data could still be sold, analyzed, or accessed by multiple entities without your explicit consent.

‍

‍How does the U.S. government’s access to data compare to China’s?

• Legal framework - In the U.S., government access generally requires court orders or warrants under the Fourth Amendment. In China, the government has broader authority and can demand access without the same legal process(Stanford DigiChina).
• Transparency and oversight - U.S. companies can disclose when data is requested, and judicial oversight is required. In China, companies are often prohibited from disclosing government data requests, and oversight is minimal(Lawfare).
• Scope of access - While the U.S. has exceptions like the PATRIOT Act for national security cases, data access is generally more limited in scope. In China, data localization laws ensure the government has broad access to stored information(Carnegie Endowment).
• Recent developments - The U.S. is restricting data flows to "countries of concern," including China, while China has implemented security assessments for data leaving the country(ITIF).

While both governments have increasing control over digital data, the U.S. offers more legal protections and oversight, whereas China’s government has direct, broad access to data stored within its borders.

‍The verdict on DeepSeek’s privacy

‍When compared to industry-leading AI models, DeepSeek AI is one of the least secure in terms of data privacy. If you value keeping your personal data safe, proceed with extreme caution - or consider safer AI alternatives that have clearer, more protective privacy policies.

‍Security vulnerabilities - how safe is DeepSeek AI?

‍Beyond privacy concerns, DeepSeek AI has demonstrated severe security vulnerabilities, making it one of the least secure AI models on the market. Recent studies have exposed critical weaknesses that make it highly susceptible to manipulation, cyber threats, and unauthorized access.

‍DeepSeek AI’s failure in security tests

‍Security researchers have conducted multiple safety tests on DeepSeek AI, and the results are concerning:

• 100% jailbreak success rate - DeepSeek AI failed to block a single harmful prompt in Cisco’s safety tests(TechStory).
• Compared to OpenAI’s GPT-4 (which blocked 86% of jailbreak attempts) and Google’s Gemini 1.5 Pro (which blocked 65%), DeepSeek’s failure rate is unprecedented(Cisco Blogs).
• DeepSeek AI was 11 times more likely to generate harmful content than OpenAI’s latest model(Euronews).

Infrastructure weaknesses

‍Beyond the AI model itself, DeepSeek’s entire infrastructure has major security flaws, leaving it vulnerable to cyberattacks.

• Over 30 publicly exposed servers were discovered, including development instances(AccuKnox).
• A ClickHouse database was accessible without authentication, allowing unrestricted access(HiddenLayer).
• Internal system metadata, chat logs, and API keys were leaked, making it easy for hackers to exploit(Qualys Blog).

Cybersecurity risks

‍DeepSeek AI is also a major cybersecurity risk due to its high susceptibility to generating insecure or malicious code.

• 78% of cybersecurity tests successfully tricked DeepSeek AI into writing insecure code or malware(Euronews).
• The model was four times more likely to generate insecure code than OpenAI’s models(GlobeNewswire).

These security flaws put users at risk, as hackers can exploit DeepSeek AI’s weaknesses to steal sensitive data, create malicious code, or spread misinformation.

‍Ethical and safety concerns - how dangerous is DeepSeek AI’s output?

‍Even beyond privacy and security, DeepSeek AI has major ethical concerns, particularly in the content it generates.

‍Harmful content generation

• DeepSeek AI was 11 times more likely to generate harmful output compared to OpenAI’s models(Euronews).
• 45% of safety tests resulted in DeepSeek producing criminal planning guides, illegal weapons instructions, and extremist propaganda(GlobeNewswire).

Bias and discrimination

• 83% of bias tests resulted in discriminatory output, showing significant racial, gender, and religious biases(GlobeNewswire).

CBRN content

• DeepSeek AI was 3.5 times more likely to produce Chemical, Biological, Radiological, and Nuclear (CBRN) content than OpenAI’s models(Euronews).

These findings show that DeepSeek AI is not just unsafe - it is actively generating dangerous content at an unprecedented scale.

Conclusion - should you use DeepSeek AI?

Why DeepSeek AI is unsafe

‍Based on the available research, DeepSeek AI is not safe for most users due to:

• Severe privacy risks with extensive data collection and storage in China.
• Major security vulnerabilities with exposed servers, leaked data, and easy jailbreaks.
• High risk of data exploitation for industrial espionage and cybersecurity threats.
• Ethical concerns, including dangerous, biased, and harmful content generation.

Final recommendation

‍Until DeepSeek AI improves its security, transparency, and compliance with international data protection laws, users should avoid it - especially if they are handling sensitive personal or corporate information. If privacy and security matter to you, consider using AI models with stronger protections:

• OpenAI’s GPT-4 – Industry-leading AI with advanced safety measures
• Google’s Gemini – A privacy-conscious AI with better security protocols
• Anthropic Claude – Designed for ethical AI interactions

OR...

Want to Use DeepSeek AI Safely Without Compromising Your Data?

If you're interested in using DeepSeek AI but don’t want to risk your privacy, there’s a solution - running it locally on your own machine. By setting up DeepSeek AI offline (for free), you can:

• Avoid data collection by keeping interactions private
• Prevent unauthorized access with full control over security
• Use AI freely without worrying about surveillance or third-party sharing

Learn how to run DeepSeek AI locally and take back control of your data. Read the full guide here: How to setup DeepSeek locally