A side event for AI Engineer Europe. Lightning talks and conversation about what it takes to ship agents that hold up in production.

Hosted by Sondera.

Day 2 of AIE Europe covers harnesses, context engineering, evals, and observability. This is the evening version of that conversation, focused on the security and governance side. If you spent the day thinking about how to build agents, come spend the evening thinking about how to keep them from breaking things.

You don't need an AIE Europe ticket to attend. If you're building agents in London or passing through for the conference, you're welcome.

Drinks and food provided.

The evening

19:30 — Doors open, less than a five-minute walk from the conference venue. Food and drinks.

20:00 — Lightning talks

Governing coding agents with policy enforcement What happens when you treat agent permissions as code instead of trusting the model to police itself. A look at deterministic controls for tool use, file access, and data flow in coding agents like Claude Code and Cursor. Matt Maisel, Co-founder & CTO, Sondera

[Second talk TBD. If you're building something relevant and want to present, reach out.]

21:15 — Open networking

22:30 — Close

Topics on the table

— Policy enforcement for agent tool calls: deterministic vs. probabilistic, and what survives an audit

— Data flow control: preventing sensitive information from leaking through agent actions

— Threat modelling for multi-agent systems: prompt injection, privilege escalation, tool misuse — What the EU AI Act, ISO 42001, and NIST AI RMF actually mean for teams shipping agents today

About Sondera

Sondera builds the security control plane for AI agents — deterministic, action-layer governance so teams can ship agents without losing sleep.

sondera.ai

Agents After Hours: Securing What You Ship