DeepSeek-R1, an AI model developed in China, is facing intense scrutiny following a study by cybersecurity firm Enkrypt AI, which found it to be 11 times more vulnerable to cybercriminal exploitation compared to other AI models. The research highlights significant security risks, including the AI’s susceptibility to generating harmful content and being manipulated for illicit activities.
This concern is further amplified by a recent data breach that exposed over a million records, raising alarms about the model’s safety. Since its launch on January 20, DeepSeek has gained immense popularity, attracting 12 million users in just two days—surpassing ChatGPT’s early adoption rate. However, its rapid rise has also triggered widespread privacy and security concerns, leading multiple governments to launch investigations or impose restrictions on its usage.
Enkrypt AI’s security assessment revealed that DeepSeek-R1 is highly prone to manipulation, with 45% of safety tests bypassing its security mechanisms. The study found that the model could generate instructions for criminal activities, illegal weapon creation, and extremist propaganda.
Even more concerning, cybersecurity evaluations showed that DeepSeek-R1 failed in 78% of security tests, successfully generating malicious code, including malware and trojans. Compared to OpenAI’s models, DeepSeek-R1 was 4.5 times more likely to be exploited for hacking and cybercrime.
Sahil Agarwal, CEO of Enkrypt AI, emphasized the urgent need for stronger safety measures and continuous monitoring to mitigate these threats.
Due to these security concerns, several countries have initiated regulatory actions.
Italy was the first to launch an investigation into DeepSeek’s privacy and security risks, followed by France, Germany, the Netherlands, Luxembourg, and Portugal. Taiwan has prohibited government agencies from using the AI, while South Korea has opened a formal inquiry into its data security practices.
The United States is also responding aggressively, with NASA banning DeepSeek from federal devices. Additionally, lawmakers are considering legislation that could impose severe fines and even jail time for those using the platform in the country.
The growing concerns surrounding DeepSeek-R1 come amid increasing competition between the US and China in AI development.
Both nations are pushing the boundaries of AI for military, economic, and technological dominance. However, Enkrypt AI’s findings suggest that DeepSeek-R1’s vulnerabilities could make it a dangerous tool for cybercriminals, disinformation campaigns, and even biochemical warfare threats. With regulatory scrutiny intensifying worldwide, the AI’s future remains uncertain as authorities weigh the risks associated with its use.
