What Could Go Wrong? Threat Modeling AI Before It Ships
How can security risks be identified before AI applications go live? Threat modeling through a practical agent example.
View sessionA prompt is a request, not a guarantee. In this hands-on session, participants try live to push a maintenance agent into impermissible answers – from bypassing a safety door to invented measurements.
A prompt is a request, not a guarantee. In this hands-on session, participants try live to push a maintenance agent into impermissible answers – from bypassing a safety door to invented measurements. We show how acceptance criteria from vehicle engineering are applied to AI agents: rules technically enforced, automatically tested and documented in a test report.
How can security risks be identified before AI applications go live? Threat modeling through a practical agent example.
View sessionPhishing, overlooked vulnerabilities and failed backups: learn from real incidents how companies end up in crisis.
View sessionWhat happens when an AI agent can use internal data and tools? Understand the risks, then put them to the test in “Roast My Agent”.
View session