Elias Botterli Sørensen
Lightning Talk
Our industry is currently excited to put large language models into everything, everywhere, all at once. As the big headline of 2025, we have put LLMs into a virtual box and called them AI agents. These agents can autonomously do stuff in our environment, without human involvement, and use the model context protocol (MCP) to access tools. This development has some serious implications for security and safety, and we have some work to do to avoid a future where hackers are the ones laughing. Get an overview of how some of this technology actually works, how it fails and why it’s a big challenge to secure it.
