Summary
- A new research study by Enkrypt AI finds that DeepSeek is 11 times more likely to generate harmful content.
- The research highlights several significant security vulnerabilities DeepSeek potentially poses.
- Governments around the globe are taking action on DeepSeek.
You may want to think twice before trying out the latest AI app, DeepSeek.
According to new research by Enkrypt AI, DeepSeek’s R1 AI model is 11 times more likely to generate harmful content that criminals could exploit than OpenAI’s o1 model. The research further says it is “highly vulnerable” to “dangerous, violent, or hateful content.”
DeepSeek made headlines recently when it became the most downloaded app on the App Store, and its R1 AI model sent shockwaves through US tech markets after showcasing its high efficiency compared to other more expensive AI models from tech giants like OpenAI and Google.
“DeepSeek-R1 offers significant cost advantages in AI deployment, but these come with serious risks. Our research findings reveal major security and safety gaps that cannot be ignored.” said Sahil Agarwal, CEO of Enkrypt AI in a press release. “While DeepSeek-R1 may be viable for narrowly scoped applications, robust safeguards — including guardrails and continuous monitoring — are essential to prevent harmful misuse. AI safety must evolve alongside innovation, not as an afterthought.”
Related
I downloaded Deepseek to see how censored it really is
Don’t expect it to be fully honest with politically sensitive topics.
Research finds DeepSeek poses cybersecurity risks
Governments across the globe are taking action on DeepSeek
EnKryptAI’s research also evaluated DeepSeek’s potential cybersecurity vulnerabilities. It found that 78 percent of the time, it could be tricked into generating malicious code, such as computer viruses. It also discovered that DeepSeek could be manipulated to create “functional hacking tools.” Both of these vulnerabilities pose a significant security risk that cybercriminals could exploit.
“Our findings reveal that DeepSeek-R1’s security vulnerabilities could be turned into a dangerous tool—one that cybercriminals, disinformation networks, and even those with biochemical warfare ambitions could exploit. These risks demand immediate attention,” Agarwal said.
Since coming into the spotlight, DeepSeek has been subject to intense scrutiny from governments worldwide. Taiwan and Australia have banned DeepSeek on all government devices, and Italy recently blocked its AI chatbot, citing concerns over its privacy policy and how it handles users’ data.
The US federal government is taking some action. NASA has banned the use of DeepSeek by federal agency employees and blocked the platform. US Senator Josh Hawley (R-MO) recently introduced a bill to ban DeepSeek and any other Chinese artificial intelligence platforms for all Americans.
Related
I downloaded Deepseek to see how censored it really is
Don’t expect it to be fully honest with politically sensitive topics.