video1.0<iframe src="https://www.loom.com/embed/5b28a99aa00a42efbe40122af2160dd9" frameborder="0" width="1920" height="1440" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>14401920Loomhttps://www.loom.com14401920https://cdn.loom.com/sessions/thumbnails/5b28a99aa00a42efbe40122af2160dd9-f6d4bb6f134bfb70.gif181.536In this video, I discuss the challenges of prompt injection in AI, particularly with OpenAI's models. I've developed a desktop app that automates the creation of Gmail and calendar events to test for these injection vulnerabilities. The app uses recent data from web searches to identify potential injection attempts and evaluates responses from OpenAI to see if they are successful. I encourage you to check out the results and provide feedback on the effectiveness of the app. Your insights will be valuable as we continue to refine our approach to this issue.