Is Cheaper Also Weaker?

DeepSeek, with a new frontier reasoning model, came like a storm and shook up the AI and LLM space in an abrupt and big way. Now that the dust is settling, it's a good time to understand its security specs. Many experts and reports have pointed out some new question marks around hardcoded keys, data control, weak encryption, SQL risks, information exposure and jail-breaks.

- By Pratima H

Open source software does not necessarily have to be a risk, but employees sharing enterprise data, APIs opening new doors and backdoors, and companies relying too soon on very-nascent technologies are all areas that enterprises should be wary of as they embrace faster and cheaper AI options shaking up the market. Joshua Goldfarb, Field CISO, F5, decodes these concerns

What are the security concerns of DeepSeek, especially those enterprises should be aware of?

While there may be many security concerns with DeepSeek, there are two primary concerns I see for enterprises. First, before enterprises use any new technology, it should be evaluated and tested so that the risk it may introduce to the enterprise is well understood. That risk can either be accepted or it can be mitigated through controls or by prohibiting the use of the technology entirely. Second, as with many technologies, DeepSeek invites employees of enterprises to share data. This opens up the potential for sensitive, confidential, and/or proprietary information to find its way out of the enterprise, which is problematic.