I saw a really interesting post from kelacyber about some early red teaming work against Deepseek. The question of how China published Deepseek at a fraction of the cost of other AI models is a huge focus right now. As a security minded person, my first concern on what was cut back during development is security. How safe is Deepseek compared to more expensive made models? Everything about Deepseek is coming at us all at rapid speed however, first signs of Red teaming is showing some scary results. For example, vulnerabilities that were patched by OpenAI are working with Deepseek.
Im sure there will be tons of red team publications coming over the next few weeks, however, check out this blog post HERE to see one example of early pen testing of Deepseek.