欢迎来到我的演讲！

First, let's take a look at a few models from organizations like DeepSeek, Qwen, and Moonshot. What are open weights? Are there any models we should try to avoid?

Next, we take no chances and create a totally sterile environment that allows us to run these LLMs on our private data. No third party APIs and no connecting to the cloud means no back doors... right?

Finally, we explore the results of some meticulous prompt engineering to uncover potential biases or even malicious responses. Are mission critical apps safe with this technology?

Thanks for joining!