AI & Beers | Securing Agentic AI: Before Your Agent Betrays You

AI agents aren't chatbots! They decide, they act, they call APIs, they handle data and when something goes wrong, you find out days later, if at all. Over the past twelve months, production agents have leaked credentials, executed unauthorized transactions, and exfiltrated confidential data through prompt injection. Classical security and QA playbooks don't catch these attack surfaces, because the trust boundary just moved, and most teams haven't caught up yet.

We're gathering for an honest conversation - no theory, no vendor pitches. On the panel, practitioners who work on agent security every day, from people leading international open-source security frameworks to QA veterans who know first-hand why deterministic testing collapses in the face of non-deterministic AI. We'll talk about the new attack surfaces, real production incidents, and what every builder needs to know before shipping agents. 

Most AI security conversations happen behind closed doors. This one happens under the open Athens sky — accessible to all levels, beer in hand, in the city that invented public discourse.

Agenda

18:45 – 19:00: Arrival & welcome

19:00 – 19:30: Panel Discussion: Securing the Agentic AI

An honest conversation about the security challenges of agentic AI — what actually goes wrong in practice, what the new attack surfaces look like, and how you build AI agents that won't betray you when incidents hit. Accessible to all levels — no prior security background required.

19:30 – 20:00: Fireside Chat — Success Stories

Honest, first-hand accounts from teams inside Greek enterprises who deployed agentic AI into production and lived through what came next. They'll walk us through the risks they actually faced.

20:00 – 20:15: Q&A — open questions from the audience 

20:15 – 21:30: Networking & Drinks