DeepSeek Security Testers Uncover ‘ Pandora’s Box’ of Cyberthreats

A security firm claims that , a startup AI model, poses serious risks in its analysis.

Calling the artificial intelligence ( AI ) model a&nbsp,” Pandora’s box”&nbsp, of risks, &nbsp, &nbsp, on Tuesday ( Feb. 11 ) from tests the company said showed a series of failures.

” The DeepSeek-R1 design underwent rigorous testing using AppSOC’s AI Security Platform”, the organization wrote on its website. ” Through a combination of automatic dynamic analysis, dynamic testing, and red-teaming techniques, the model was put through scenarios that resemble real-world attacks and security stress tests. The effects were alarming”.

For instance, the testing revealed that DeepSeek made it simple for people to create malware and viruses. The model displayed a 98.8 % failure rate when testers asked it to create malware, and an 86.7 % failure rate when asked to create virus code.

In contrast, the design showed a 68 % failure rate when prompted to create “responses with dangerous or harmful language, indicating bad safeguards”, AppSOC said. It even produced&nbsp, &nbsp, — technically incorrect or fabricated data — 81 % of the time.

Mali Gorantla, AppSOC’s co-founder and main scholar, told the site Dark Reading that this functionality shows that — in spite of the hype around DeepSeek’s lower price and open source model — companies should avoid it in its latest version.

” For most business applications, failure rates of about 2 % are considered unacceptable”, he said. Our suggestion is to prevent anyone from using this design for any business-related AI.

PYMNTS has reached out to DeepSeek for reply, but they have not already received a response.

With the launch of its newest AI design next month, China-based DeepSeek shocked the tech industry. Some people claimed that the cost of its development indicated that creating AI instruments similar to OpenAI’s ChatGPT was affordable.

However, the company’s design has since come under fire in the U. S., with officials&nbsp, &nbsp, that would keep it off government-owned equipment.

However, White House AI king David Sacks told Fox News last quarter that it is&nbsp,” significant information” &nbsp, that DeepSeek used OpenAI’s designs to create its own technology.

And Google’s AI captain, &nbsp, , said earlier this year the idea that DeepSeek spent just under$ 6 million to build an AI unit that can compete with those of American tech giants is&nbsp, “”.

He claimed that DeepSeek” seems to have only reported the cost of the final training round, which is only a small portion of the total cost.”