Trail AI, Risky Prompts
Omer Frank designed Risky Prompts in Trail AI, the investigation flow that takes security teams from a risky prompt alert to a clear view of what was asked, how the model responded and what data or actions were at risk (2024).
Trail AI, Risky Prompts is part of an AI-powered DLP platform, focused on uncovering dangerous sessions triggered by sensitive or malicious conversations. I designed the investigation flow and interface that takes security teams from a high-level alert, risky prompt detected, into a clear, navigable view of what was asked, how the model responded, and what data or actions were at risk.
The experience groups prompts by risk type, highlights patterns across sessions, and gives analysts guided next steps so they can quickly decide whether to escalate, remediate, or tune policies, without having to wade through raw logs or opaque AI behavior.
All 2 screens