About the Speaker
< Talk Title />
< Talk Category />
< Talk Abstract />
LLMs are omnipresent and being bolted onto almost every piece of Software nowadays. While convincing a web shop’s AI chat bot to give you bomb building instructions might be a fun exercise, the real problems start to arise when we let those bots do things on our behalf.
Brace yourself for some horror stories about prompts, injections and AI (double) agents meeting classic application security issues. But also let us have a look at safeguards, mitigations and potential ways to not let a vibe coding session ruin your production database completely.
< Speaker Bio />
Jörn, in most places in the Internet known by his nickname joernchen, is a Principal Security Engineer in GitLab’s Security Research Team. Since the LLMs first went mainstream, he’s found himself diving deeper into AI security - driven by his work securing cutting-edge AI features for customers and by the lovely smell of fresh attack surface manifesting.