Research

Security research for the systems that think, fly, drive and sail.

Software now runs far more than websites. XCODITS studies how attackers reach AI models, drones, vehicles and vessels, and turns what we learn into defences that people can actually use.

01 · AI security

Keeping AI models and agents on your side.

Businesses are handing real decisions to AI. We study how those systems can be tricked, leak data or be turned against the people who rely on them.

Prompt injection

Hidden instructions in documents, emails and web pages that make an AI agent act against its owner.

Data leakage

How models expose training data, private records or internal tools to people who should not see them.

Model supply chain

Tampered models, datasets and plugins that arrive with a backdoor already built in.

Safe agent design

Permissions, review steps and monitoring that keep autonomous agents inside their limits.

From lab to production

Every finding feeds the tools and teams that protect you.


Working on something in one of these fields?

Whether you build it, run it or need it tested, tell us what you are working on and a person will reply within one working day.