top of page

Is It Safe To Use DeepSeek AI ?

OCM Engineers

DeepSeek is a Chinese artificial intelligence (AI) company founded in May 2023.


In April 2023, High-Flyer initiated an artificial general intelligence lab dedicated to developing AI tools separate from its financial operations. This lab evolved into DeepSeek, focusing on creating open-source large language models (LLMs). The company's flagship model, DeepSeek-R1, was is claimed to have been developed at a fraction of the cost of comparable models such as ChatGTP and Gemini.


DeepSeek AI logo supporting a blog post " is DeepSeek Safe"

By January 2025, DeepSeek's chatbot application, based on the DeepSeek-R1 model, became the most-downloaded free app on the iOS App Store in the United States, this rapid ascent underscores DeepSeek's impact on the AI industry and its potential to influence future developments in artificial intelligence


#1: DeepSeek Disrupts The Established AI Order


This week, the global technology sector experienced significant volatility following the emergence of DeepSeek.


DeepSeek's release of a cost-effective AI model, developed with substantially lower resources than its western counterparts, has raised concerns about the competitive landscape in AI development. Major tech companies, particularly those heavily invested in AI, saw notable declines in their stock prices. For instance, Nvidia's shares dropped nearly 17%, marking a record one-day loss in market capitalisation. Similarly, other tech giants like Microsoft and Alphabet faced downturns. Analysts suggest that DeepSeek's advancements could democratise AI development, potentially leading to more accessible and affordable AI solutions globally.

#2: Does DeepSeek Work?


Here's a summary of DeepSeek's strengths and weaknesses:


Strengths:


  • Cost-effective:  DeepSeek is completely free to use, making it a strong competitor to paid models like ChatGPT Plus.   

  • Comparable performance:  In some benchmarks and anecdotal comparisons, DeepSeek's performance is comparable to, or even surpasses, that of ChatGPT's free model (o1) and even Google's Gemini.   

  • Efficient design: DeepSeek's developers have focused on efficiency, allowing it to achieve strong results with fewer computational resources and older hardware.   

  • Open source: The model is open source, which encourages community involvement and wider accessibility for developers.   


Weaknesses:


  • Limited features: Compared to more established chatbots, DeepSeek currently lacks some features, such as multi-modal capabilities (handling images and other data types) and plugins.   

  • Memory: DeepSeek's memory within a conversation is limited compared to ChatGPT, which can retain information from previous interactions.


    There have been accusations that DeepSeek AI has benefited from the work of established players in the field, particularly OpenAI. Some argue that DeepSeek's ability to achieve comparable performance with fewer resources and older hardware suggests they may have utilised insights or techniques from existing models.


    However, there is no concrete evidence to prove that DeepSeek has directly stolen from other companies. Although it does respond to some queries claiming to be ChatGTP ! DeepSeek's developers maintain that their model is built on independent research and innovation. They emphasise their focus on efficiency and open-source development as key factors in their success.

       

#3: Is DeepSeek Safe ? Pros and Cons


DeepSeek's rise has been accompanied by significant security and privacy concerns.


Data Privacy: DeepSeek's privacy policy indicates that user data, including text inputs, chat history, and uploaded files, are stored on servers located in China.


Security Vulnerabilities: A recent investigation revealed that DeepSeek left a critical database exposed on the internet, leaking over one million records, including system logs, user prompts, and API authentication tokens. This exposure underscores potential security immaturity and poses risks to user data.


Censorship and Content Control: DeepSeek has been observed to employ censorship mechanisms, particularly on topics deemed politically sensitive by the Chinese government. For instance, the chatbot refuses to answer questions about events like the 1989 Tiananmen Square protests, human rights in China, and other sensitive subjects, often responding with messages like, "Sorry, that's beyond my current scope. Let's talk about something else."


Official Responses: In light of these concerns, various entities have taken precautionary measures. The U.S. Navy has advised its personnel against using DeepSeek, citing potential security and ethical issues. Similarly, the Australian government is evaluating the app's safety, with officials indicating that a ban could be implemented if security agencies deem it unsafe.


Given these factors, both individuals and businesses should exercise caution when considering the use of DeepSeek. It's essential to be aware of the potential risks associated with data privacy, security vulnerabilities, and content censorship inherent in the application.


#4: What About Privacy?


Is DeepSeek safe ? We don't know is the answer and because we do not know, exercise caution.


Deepseek is open source which means its code is available for scrutiny and change. Generally this is a good thing as experts can scrutinise the code for malicious actions and vulnerabilities.


The DeepSeek app is available for Apple and Android platforms and to appear in the respective stores there must be controls implemented as to data access and privacy on your devices.


However, DeepSeek's privacy policy indicates that user data, including text inputs, chat history, and uploaded files, is stored on servers located in China. This raises concerns about potential access by the Chinese government, as companies operating in China are subject to national cyber security laws that may require data sharing with authorities upon request.


Here's a breakdown of DeepSeek's safety considerations for individuals and businesses:


For Individuals:


  • Data Privacy: DeepSeek's privacy policy is not as clear as some of the bigger players like OpenAI or Google. It's unclear where your data is stored and how it's used. If you're concerned about data privacy, you might want to be cautious about what information you share with DeepSeek.   

  • Unexpected Outputs: Like all AI chatbots, DeepSeek can sometimes produce unexpected or inaccurate outputs. It's important to critically evaluate the information it provides and not rely on it for making important decisions.

  • Bias and Offensive Content: AI models are trained on massive datasets of text and code, which can contain biases and offensive content. While DeepSeek likely has safeguards in place, it's still possible for it to generate biased or offensive outputs.   


For Businesses:


  • Data Security: Businesses should be particularly cautious about sharing sensitive information with DeepSeek, especially given the lack of clarity around its data privacy practices.

  • Reliability and Accuracy: DeepSeek is still a relatively new model, and its reliability and accuracy may not be as well-established as that of more mature AI systems. Businesses should be aware of this when considering using DeepSeek for critical tasks.

  • Ethical Considerations: Businesses should consider the ethical implications of using AI, including the potential for bias and discrimination. It's important to use AI responsibly and ensure that it aligns with the company's values and ethical guidelines.


Conclusion


DeepSeek can be a powerful tool for both individuals and businesses, but it's important to use it with caution and awareness of its potential risks. If you're concerned about data privacy or security, you may want to wait for more information about DeepSeek's practices or consider using a more established AI provider.


Perplexity has introduced the Deepseek R1 model as it is open-source and accessible to anyone interested in using it, addressing two main concerns: firstly, the possibility of China censoring the results provided by DeepSeek, and secondly, the risk of the model transmitting user data to the Chinese government.



OCM suggest that Perplexity is a safer starting point for evaluation of DeepSeek than using the model directly.








13 views0 comments

Comments


Commenting has been turned off.
bottom of page