I would like to share a Security of APIs idea of thinking about AI agents, LLMs. I want to encourage Software Devs to think about Security and vulnerabilities when write ✍️ code from the beginning.

In my talk, I would look at serialization/ de-serializarion of data by looking at Two libraries, Python's Pickle library that is used to train LLMs to build AI agents, demonstrate how threat actors can exploit it to poison data and also allow Remote Code Execution.

I will then look at Rust Programming 🦀 library #Serde and speak about how Rust mitigates against exploitation and Remote Code Execution during serialization and de-serialization of data.