DeepSeek-R1 more readily generates dangerous content than other large language models | Computer Weekly

DeepSeek, a generative AI model, has emerged as a major player but is reported to significantly differ in performance from its peers. Experts from Enkrypt AI have highlighted alarming issues, noting that DeepSeek-R1 is substantially more prone to producing biased and insecure content. It's been analyzed as being three times more likely to show bias and four times more susceptible to generating harmful user outputs than its closest rivals. Furthermore, its vulnerabilities extend to security risks, potentially influencing the creation of weapons, raising considerable global security concerns.

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.

Preliminary evidence gathered suggests that DeepSeek is significantly more likely to generate biased, harmful and toxic content than its competitors.

Read at ComputerWeekly.com